Suppression of CCL2 angiocrine function by adrenomedullin promotes tumor growth

The authors identify a tumor-suppressing angiocrine function of CCL2 which is inhibited by tumor-derived adrenomedullin acting via its receptor on endothelial cells. Disinhibition of endothelial CCL2 results in reduced tumor progression through inhibition of adrenomedullin formation by tumor cells.


Introduction
Tumor cells and endothelial cells interact in multiple ways within the tumor microenvironment, which critically contributes to tumor progression, metastatic dissemination, and immune surveillance of cancer cells, as well as to the response to therapies (De Palma et al., 2017;Maman and Witz, 2018;Nagl et al., 2020). It is well established that both tumor cells and the tumor microenvironment regulate endothelial cells through various pro-and anti-angiogenic factors. However, within the tumor stroma, endothelial cells not only respond to tumorderived angiogenic factors but have also been recognized during the last decade to produce and release factors that regulate neighboring cells including tumor cells. These angiocrine factors include growth factors, interleukins, or chemokines, and have been shown to critically contribute to cancer progression and metastasis (Butler et al., 2010;Maishi et al., 2019).
Adrenomedullin is a regulatory peptide that was first isolated from human adrenal tumor (Kitamura et al., 1993). Multiple tumors have been shown to express adrenomedullin, and it is believed to function as a potent autocrine or paracrine factor that promotes tumor cell survival and lymphangiogenesis (Klein and Caron, 2015;Larrayoz et al., 2014;Zudaire et al., 2003).
Adrenomedullin has also been proposed as a mediator of tumor angiogenesis (Nikitenko et al., 2006). This is based on the ability of adrenomedullin to induce pro-angiogenic effects in endothelial cells in vitro (Kim et al., 2003;Martinez et al., 2002;Zhang et al., 2017) as well as on the observation that the expression levels of adrenomedullin in tumors correlate with tumor angiogenesis in vivo (Hague et al., 2000;Martinez et al., 2002;Oehler et al., 2002). However, evidence for a direct proangiogenic effect of adrenomedullin in tumors is missing and how adrenomedullin activity is regulated in tumors is poorly understood.
Adrenomedullin exerts its cellular effects by activating the G-protein-coupled receptor calcitonin receptor-like receptor (CALCRL), which requires the presence of receptor activitymodifying proteins 2 or 3 (RAMP2 or RAMP3) to function as an adrenomedullin receptor (Poyner et al., 2002;Schonauer et al., 2017;Vazquez et al., 2020). The adrenomedullin receptor is coupled to the G-protein G s which activates adenylyl cyclase resulting in increased formation of cAMP, a pathway which has been shown to increase endothelial barrier function and nitric oxide formation and to mediate anti-inflammatory effects in endothelial cells (Iring et al., 2019;Moy et al., 1998;Nakayama et al., 2020).
Based on co-culture experiments in vitro as well as on in vivo experiments, we show that tumor cell-derived adrenomedullin exerts a pro-angiogenic effect on endothelial cells, which is mediated by the G s -coupled adrenomedullin receptor. In addition, we show that adrenomedullin acting on endothelial cells further stimulates formation of adrenomedullin in tumor cells by inhibiting the endothelial formation and release of CCL2, which acts as an angiocrine factor to inhibit adrenomedullin expression by tumor cells. Increasing endothelial CCL2 formation may therefore be a strategy to decrease tumor growth.

Results
Loss of endothelial G s -mediated signaling reduces tumor progression To investigate the role of endothelial G s -mediated signaling in vivo, we analyzed mice with tamoxifen-inducible endotheliumspecific deletion of Gα s , the α-subunit of the G-protein G s (Tek-CreERT2;Gnas flox/flox , herein referred to as EC-Gα s -KO mice; Iring et al., 2019). Loss of endothelial Gα s had no effect on the development of the retinal vasculature at postnatal day 7 (P7; Fig. S1 a). However, when we examined endothelial Gα s function in syngeneic tumor models, we found that growth of B16-F10 melanoma was significantly decreased in EC-Gα s -KO mice compared to control animals ( Fig. 1 a). Tumors in EC-Gα s -KO mice showed reduced proliferation of endothelial and nonendothelial cells, reduced vessel density, reduced vessel coverage, increased cell death, and an increased hypoxic area ( Fig. 1 b and Fig. S1 b). Lewis lung carcinoma (LLC1), when grown in EC-Gα s -KO mice, also showed reduced growth as well as reduced cell proliferation, reduced vessel density without affecting lymphatic vessel area, and an increased hypoxic area ( Fig. 1 c and Fig. S1, c and d), whereas immune cell infiltration was comparable between control and EC-Gα s -KO mice (Fig. S1, e and f). We then studied a spontaneous breast cancer model after crossing control or EC-Gα s -KO mice with a mouse line in which the polyoma middle T-antigen expression is driven by the mouse mammary tumor virus promoter (MMTV-PyMT). Endothelial loss of Gα s delayed tumor growth in MMTV-PyMT mice ( Fig. 1 d), and this effect was accompanied by reduced vessel density and reduced proliferation of endothelial and non-endothelial cells (Fig. 1 e). These data indicate that endothelial Gα s plays a critical role in tumor growth.
The adrenomedullin receptor CALCRL as well as G s are required for endothelial and tumor cell proliferation To understand the role of endothelial Gα s in tumor progression, we performed an siRNA-mediated knock-down of Gα s in HUVECs and cultured the cells with or without GFP-expressing MeWo human melanoma or MDA-MB-231 human breast cancer cells. Gα s knock-down in HUVECs did not have a significant effect on the proliferation of endothelial cells grown in the absence of tumor cells (Fig. 2 a and Fig. S2 a). However, when HUVECs were cultured together with MeWo or MDA-MB-231 cells, the number of proliferating endothelial cells increased, and this effect was lost after siRNA-mediated suppression of endothelial expression of Gα s (Fig. 2 a and Fig. S2 a). Interestingly, knockdown of Gα s also strongly reduced tumor cell proliferation in this co-culture model (Fig. 2,b and c;and Fig. S2 b), an effect which was not due to a change in the rate of apoptosis of endothelial cells or tumor cells (Fig. 2 d and Fig. S2 c). These results indicate that endothelial G s -mediated signaling promotes endothelial cell proliferation only in the presence of tumor cells and that it also increases proliferation of tumor cells.
To identify the receptor operating upstream of G s in endothelial cells, we performed an siRNA-mediated knock-down of 23 G s -coupled receptors or orphan receptors which are highly expressed in HUVECs and primary mouse lung endothelial cells (MLECs; Fig. S2 d), and determined endothelial cell proliferation in the absence or presence of MeWo melanoma cells (Fig. 3 a and Fig. S2 e). While suppression of the majority of the candidate receptors had no effect on MeWo-dependent endothelial proliferation, knock-down of the CALCRL as well as of the adhesion GPCR latrophilin-2 (ADGRL2) inhibited MeWo-induced proliferation of endothelial cells to a similar degree as knock-down of Gα s (Fig. 3 a). Since knock-down of ADGRL2 also reduced basal proliferation of HUVECs in the absence of MeWo cells (Fig. S2 e), we focused on CALCRL, which functions as a receptor for adrenomedullin (Poyner et al., 2002). Similar to knock-down of endothelial Gα s , loss of endothelial CALCRL or blockade of CALCRL by the competitive adrenomedullin receptor antagonist adrenomedullin 22-52 (AM22-52) reduced not only the tumor cell-induced growth of endothelial cells (Fig. 3, b and c) but also tumor cell proliferation in the co-culture model (Fig. 3, d-f).
To test the role of endothelial CALCRL in vivo, we analyzed tamoxifen-inducible endothelium-specific CALCRL-deficient mice (Tek-CreERT2;Calcrl flox/flox , herein referred to as EC-Calcrl-KO mice; Iring et al., 2019). Similar to EC-Gα s -KO mice, loss of endothelial CALCRL had no effect on retinal vasculature development (Fig. S2 f). However, tumor growth of subcutaneously injected B16-F10 and LLC1 cells as well as orthotopically injected E0771 breast cancer cells was significantly decreased in EC-Calcrl-KO mice compared to control animals (Fig. 3,g and h;and Fig. S2 g). Tumors in EC-Calcrl-KO mice showed reduced endothelial and nonendothelial cell proliferation as well as reduced vessel density compared to control animals (Fig. 3 i). These data indicate that the G s -coupled adrenomedullin receptor on endothelial cells mediates endothelial and tumor cell proliferation in vitro and in vivo.
Tumor cell-derived adrenomedullin promotes endothelial and tumor cell proliferation Adrenomedullin is expressed both in endothelial cells and tumor cells and can induce the proliferation of both endothelial cells and tumor cells (Hinson et al., 2000;Nikitenko et al., 2006) in a CALCRL-and Gα s -dependent manner (Fig. 4, a and b). To identify the major source of adrenomedullin responsible for the proliferation of endothelial and tumor cells in the co-culture model, we studied the effect of an adrenomedullin knockdown in endothelial or in tumor cells on the proliferation of both co-cultured cell types. Suppression of adrenomedullin expression in HUVECs had no effect on the proliferation of endothelial cells when cultured alone (Fig. 4 a) or when co-cultured with tumor cells (Fig. 4 c and Fig. S3 a). Also, tumor cell proliferation was unaffected by suppressed endothelial ADM expression ( Fig. 4 d and Fig. S3 b). In contrast, suppression of adrenomedullin expression in tumor cells led to a strong reduction in the proliferation of HUVECs (Fig. 4 e and Fig. S3 c) and of co-cultured tumor cells (Fig. 4 f). However, loss of adrenomedullin expression had no effect on growth of tumor cells cultured alone (Fig. 4 b). In addition, suppression of adrenomedullin expression in MeWo melanoma cells, but not in HUVECs, led to a strong reduction in adrenomedullin levels in the co-culture (Fig. S3 d). These data indicate that it is primarily tumor cell-derived adrenomedullin that promotes endothelial and tumor cell proliferation in the co-culture model.
Similar observations were made in a murine in vitro coculture model. When primary MLECs were cultured together with B16-F10 cells, the proliferation of endothelial and tumor cells was strongly reduced after knock-down of endothelial Gα s (Fig. 5, a and b). In addition, suppression of adrenomedullin expression in tumor cells, but not in MLECs, led to reduced adrenomedullin expression and release in the co-culture (Fig. 5, c and d). Consistent with a role of tumor cell-derived adrenomedullin in endothelial and tumor cell proliferation under coculture conditions, MLECs showed increased proliferation when exposed to conditioned medium of B16-F10 cells, an effect lost after knock-down of endothelial Gα s and CALCRL (Fig. S3 e). However, conditioned medium from melanoma cells transduced  Similarly, tumors from MMTV-PyMT mice crossed with EC-Gα s -KO mice expressed significantly less adrenomedullin compared to tumors from normal MMTV-PyMT animals ( Fig.  S4 b). To test whether G s -mediated signaling in endothelial cells regulates the release of a diffusible endothelial factor, which in turn controls the formation of adrenomedullin in tumor cells, we cultured tumor cells in the presence of supernatants from endothelial cells with or without loss of Gα s . Whereas supernatants of control endothelial cells had no effect on the amount of adrenomedullin in tumor cells as well as on the release of adrenomedullin from tumor cells, supernatants from endothelial cells with suppressed Gα s expression inhibited both expression and release of adrenomedullin (Fig. 6, c and d). This strongly indicates that G s -mediated signaling in endothelial cells regulates the formation and/or release of adrenomedullin from tumor cells through a diffusible factor.
To search for potential factors controlled by G s -mediated signaling in endothelial cells, we compared the transcriptome of control HUVECs and of HUVECs lacking Gα s . We found various genes whose expression was strongly up-or down-regulated after suppression of GNAS, the gene encoding Gα s (Fig. S4 c). When we tested the effect of adrenomedullin on the expression of these genes, we found that expression of CCL2 was strongly altered (Fig. S4 d). Expression of CCL2, which encodes a chemokine belonging to the CC chemokine family, increased after knock-down of Gα s but was reduced in control cells exposed to adrenomedullin ( Fig. 6 e). Knock-down of CALCRL also increased CCL2 expression ( Fig. 6 f), and incubation of cells with dibutyryl-cAMP, a membrane-permeable stable derivative of cAMP mimicking Gα s activation, suppressed CCL2 expression induced by knock-down of CALCRL or Gα s (Fig. 6 f). The increased endothelial expression of CCL2 after inhibition of CALCRL/G s -mediated signaling was accompanied by an increased CCL2 level in cellular supernatants ( Fig. 6 g). Furthermore, Ccl2 expression in tumor ECs isolated from EC-Gα s -KO mice was higher compared to control mice ( Fig. 6 h), and tumors from MMTV-PyMT mice crossed with EC-Gα s -KO animals also expressed significantly more Ccl2 compared to tumors without loss of endothelial Gα s (Fig. S4 b). These results show that CCL2 expression in ECs is inhibited by adrenomedullin acting through its G s -coupled receptor and subsequent cAMP signaling.

Endothelial CCL2 inhibits expression of adrenomedullin by tumor cells
To test whether CCL2 inhibits adrenomedullin formation in tumor cells, we incubated different tumor cells with CCL2. CCL2 indeed induced an inhibition of adrenomedullin expression, which was not seen after knock-down of its receptor CCR2 (Fig.  S4, e and f). The effect of CCL2 on adrenomedullin expression was blocked by preincubation of cells with pertussis toxin indicating involvement of G i -type G-proteins (Fig. S4 g). We then analyzed the effect of endothelial loss of CCL2 on tumor cells and endothelial cells in the co-culture model. Knock-down of CCL2 in HUVECs with suppressed Gα s expression had no effect on endothelial cell proliferation (Fig. 7 a) but reverted the effect of the Gα s knock-down on tumor cell proliferation (Fig. 7, b and c) as well as on adrenomedullin levels (Fig. 7 d). In addition, the reduced adrenomedullin expression and proliferation of control presence of GFP-MeWo tumor cells was determined by staining for Ki67. Shown are the ranked average ratios of Ki67-positive HUVECs co-cultured with MeWo cells and Ki67-positive HUVECs grown alone (a) and representative immunofluorescence staining (b; n = 4 independently evaluated dates). (c-f) HUVECs transfected with control siRNA or siRNA directed against CALCRL (e and f) or pretreated without or with 1 µM of the adrenomedullin receptor antagonist AM22-52 (c and d) were cultured with GFP-MeWo for the indicated time periods. Thereafter, Ki67-positive HUVECs were determined by staining with anti-Ki67 antibody (red) with DAPI (blue; c, n = 3 independent experiments), the number of GFP-expressing tumor cells was determined by either immunofluorescence (d and e, n = 3 independent experiments), or the expression level of GFP was determined by Western blot analysis (f). The bar diagrams show the statistical evaluation or represent the relative densitometric values based on Fiji software (f; n = 3 independent experiments). (g) B16-F10 melanoma cells were injected subcutaneously in control or EC-Calcrl-KO mice, and tumor growth was determined (n = 7 mice for each genotype). (h) E0771 breast cancer cells were injected into mammary fat pad of control or EC-Calcrl-KO mice, and tumor growth was determined (n = 5 mice for each genotype). (i) Immunohistochemistry of B16-F10 tumors grown in control or EC-Calcrl-KO mice. Sections were analyzed for markers of endothelial cells (PECAM1; green), perivascular cells (α-SMA; red) or proliferating cells (Ki67; red), and were stained with DAPI (blue). Bar diagrams show the statistical evaluation (n = 6 mice for each genotype). Bar length: 100 μm (b-e and i). Data represent mean values ± SEM; **, P ≤ 0.01; ***, P ≤ 0.001 (two-way ANOVA and Bonferroni's post hoc test [f-h], two-tailed Student's t test [c-e and i]).  Fig. S4 h).
Finally, we analyzed the role of endothelial CCL2 in tumor progression under in vivo conditions. Endothelium-specific CCL2-deficient mice (Tek-CreERT2;Ccl2 flox/flox mice, herein referred to as EC-Ccl2-KO mice) did not show any differences in tumor growth when injected with B16-F10 melanoma cells (Fig. 8 a). However, when crossed with EC-Gα s -KO mice to generate inducible endothelium-specific Gα s /CCL2 doubledeficient mice (EC-Gα s /Ccl2-dKO) we found that loss of endothelial CCL2 normalized the tumor phenotype of EC-Gα s -KO animals ( Fig. 8 b). Tumors from EC-Gα s -KO mice showed significantly higher CCL2 levels and reduced Adm expression compared to tumors of control mice (Fig. 8 c). However, tumors of EC-Gα s /Ccl2-KO mice had normalized adrenomedullin expression compared to EC-Gα s -KO animals (Fig. 8 c). In contrast to EC-Gα s -KO mice, EC-Gα s /Ccl2-dKO animals showed increased tumor cell proliferation, whereas tumor vasculature was hardly affected (Fig. 8 d). In addition, suppression of CCR2 expression in tumor cells normalized tumor growth in EC-Gα s KO animals, indicating that tumor cells are the main effector of endothelial CCL2 (Fig. 8 e). These data indicate that endothelial CCL2 can inhibit adrenomedullin formation but that this in turn is suppressed by adrenomedullin itself, which through its G s -coupled receptor on endothelial cells attenuates CCL2 formation.

Discussion
Since endothelial cells when exposed to adrenomedullin in culture show typical angiogenic responses and, since overexpression of adrenomedullin in tumors or systemic application of adrenomedullin results in increased vascular and lymphatic angiogenesis (Iimuro et al., 2004;Karpinich et al., 2013;Martinez et al., 2002;Oehler et al., 2002;Zhang et al., 2017), adrenomedullin has been suggested to mediate tumor angiogenesis. However, evidence for a direct pro-angiogenic effect of adrenomedullin in tumors is still missing. Here, we show that loss of the adrenomedullin receptor or of the G-protein G s in endothelial cells results in strongly reduced tumor angiogenesis in different syngeneic tumor models as well as in a genetic cancer model. In addition, our data revealed a novel mechanism by which adrenomedullin promotes its formation by tumor cells. In addition to its pro-angiogenic effect, adrenomedullin also inhibits endothelial formation of CCL2, which acts as an angiocrine factor and suppresses formation of adrenomedullin by tumor cells (Fig. 8 f). This confirms the role of adrenomedullin as a pro-angiogenic factor which also directly promotes tumor growth. Other proangiogenic factors such as VEGF and FGF can also affect tumor cells. It remains to be explored to what degree these proangiogenic and tumor-promoting factors synergize in particular tumors.
Our data clearly show that the endothelial G s -coupled adrenomedullin receptor mediates pro-angiogenic effects. How G s -mediated signaling resulting in increased cAMP levels promotes tumor angiogenesis is unclear. Several other G s -coupled receptors have been shown to increase tumor angiogenesis (Balzan et al., 2012;Garg et al., 2017;Yang et al., 2013;Zahalka et al., 2017) and the cAMP-dependent protein kinase A (PKA) has been shown to be a critical regulator of tumor angiogenesis (Bir et al., 2012;Garg et al., 2017;Xue et al., 2018). During development PKA has been shown to prevent hypersprouting and to regulate the number of tip cells, an effect which involves inhibitory regulation of autophagy (Nedvetsky et al., 2016;Zhao et al., 2019). Another study has shown that the cAMP/PKA pathway can promote the release of VEGF (Garg et al., 2017) and, in prostate cancer endothelial cells, the G s -coupled β2 adrenergic receptor has been shown to induce an angiogenic switch, which occurs through the alteration of endothelial cell metabolism by inhibiting endothelial oxidative phosphorylation (Zahalka et al., 2017). In addition, it is also possible that G s /cAMP-mediated proangiogenic effects are mediated by the cAMP-activated Rap1 GTP exchange factor EPAC, which has been shown to promote endothelial responses during tumor angiogenesis (Chrzanowska-Wodnicka, 2017; Garg et al., 2017).
Whereas the loss of the adrenomedullin receptor as well as of G s in endothelial cells reduced tumor angiogenesis, postnatal angiogenesis in the retina was not affected. This may underline the functional differences between tumor and developmental angiogenesis (Chung and Ferrara, 2011;Jin and Jakobsson, 2012;Papetti and Herman, 2002) and indicates that endothelial adrenomedullin signaling pathway is not involved in developmental retinal angiogenesis. In contrast to our data, it has been reported that an endothelium-specific knock-out of RAMP2, which together with CALCRL forms the adrenomedullin receptor, affects both retinal angiogenesis and tumor angiogenesis (Iesato et al., 2013;Tanaka et al., 2016). This difference is most likely due to the requirement of RAMP2 for modulation of the function of a different G-protein-coupled receptor. In fact, RAMP2 has been shown to interact with several other GPCRs, including GPR4 and GPR182 (Lorenzen et al., 2019;Serafin et al., 2020), which are expressed in endothelial cells (Jing et al., 2016;Kechele et al., 2017;Le Mercier et al., 2021;Wyder et al., 2011).
The bar diagram shows the statistical evaluation. (b) GFP-MeWo cells were transfected with control siRNA or siRNA directed against CALCRL or ADM and were cultured without (basal) or with 2 nM adrenomedullin for 24 h. Thereafter, the number of GFP-expressing tumor cells was determined. The bar diagram shows the statistical evaluation (a and b; n = 2 independent experiments, two evaluated areas per experiments). (c and d) HUVECs were transfected with control siRNA or siRNA directed against ADM and were cultured with GFP-MeWo for the indicated time periods. Endothelial proliferation was determined by staining with anti-Ki67 antibody (c), or the expression level of GFP was determined by Western blot analysis as an indicator of tumor cell growth (d).
(e and f) HUVECs were cultured with GFP-MeWo transfected with control siRNA or siRNA directed against ADM for the indicated time periods. Endothelial proliferation was determined by staining with anti-Ki67 antibody (e), or the tumor cell proliferation was determined by quantifying expression level of GFP by Western blot analysis (f). The bar diagrams show the statistical evaluation (c and e; n = 3 independent experiments) or represent the relative densitometric values based on Fiji software (d and f; n = 3 independent experiments). Bar length: 100 μm (a-c and e). Data represent mean values ± SEM; *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001 (one-way ANOVA and Tukey's post hoc test [a and b], two-way ANOVA and Bonferroni's post hoc test [d and f], and two-tailed Students t test [c and e]).
In addition to the pro-angiogenic effect, we found that adrenomedullin also suppressed the formation of endothelial CCL2 through cAMP-PKA, and thereby resulted in the suppression of CCL2-induced inhibition of adrenomedullin formation by tumor cells. This inhibitory regulation resulted in increased adrenomedullin signaling and a self-reenforcement of endothelial adrenomedullin-induced CCL2 inhibition. Adrenomedullin has been shown to inhibit CCL2 production in endothelial cells and other cell types (Huang et al., 2011;Iwamoto et al., 2003;Yuda et al., 2012). Since cAMP has been shown in endothelial cells to inhibit NF-κB-mediated transcriptional regulation (Nakayama et al., 2020;Ollivier et al., 1996) and since NF-κB is well known to play a central role in regulating CCL2 expression (Martin et al., 1997), this would be the most likely mechanism underlying adrenomedullin-induced suppression of CCL2 expression.
The chemokine CCL2 and its receptor CCR2 have been shown to play multiple roles in tumorigenesis and tumor cell metastasis (Borsig et al., 2014;Gschwandtner et al., 2019;Zhang et al., 2010). Various data show that enhanced levels of CCL2 in different tumors promote tumor progression and metastasis. This is due to different mechanisms including the ability of CCL2 to recruit and regulate tumor-associated macrophages, which promote tumor progression and metastasis in part through its immunosuppressive role in the tumor microenvironment (Fridlender et al., 2011;Kersten et al., 2017;Li et al., 2017;Yang et al., 2019). CCL2 has also been shown to promote tumor progression by acting on endothelial cells to stimulate tumor cell extravasation and metastasis (Hauselmann et al., 2016;Wolf et al., 2012) or by a direct effect on tumor growth (Wong et al., 2020). However, CCL2 has also been shown to decrease tumor cell growth and metastasis (Asano et al., 1996;Rollins and Sunday, 1991;Takahashi et al., 2009). Several reports showed that CCL2 can synergize with bacterial endotoxin to activate macrophages to become tumoricidal and to reduce tumor progression and metastasis (Huang et al., 1995;Nakashima et al., 1998;Singh et al., 1993). In addition, CCL2 can have antitumorigenic and antimetastatic activity depending on cellular context and cancer type by regulating neutrophil activity (Granot et al., 2011;Lavender et al., 2017). Our data show a new role of CCL2 as an angiocrine factor of the tumor microenvironment which suppresses tumor progression by directly acting through CCR2 expressed by tumor cells. However, this activity can be suppressed by tumor-derived adrenomedullin and potentially also by other mediators which activate G s -coupled receptors. Consistent with this, CCL2 has been found to be one of the most strongly downregulated genes in endothelial cells of invasive breast cancer compared to endothelial cells of normal mammary vasculature (Parker et al., 2004).
In summary, our data demonstrate that adrenomedullin is an essential regulator of tumor angiogenesis which, by acting through its G s -coupled receptor on endothelial cells, promotes endothelial cell proliferation. In addition, we show that adrenomedullin inhibits the formation of the endothelial angiocrine factor CCL2 which, in turn, can suppress adrenomedullin formation by tumor cells. It remains to be evaluated whether promoting endothelial CCL2 formation in adrenomedullinproducing tumors is a strategy to reduce tumor growth.

Cells
HUVECs and MLECs were purchased from Lonza. HUVECs were cultured with EGM-2 (Lonza) and MLECs were cultured with endothelial growth medium (EGM-2-MV, Lonza). Confluent cells at passage ≤6 were used in all experiments. MDA-MB-231-GFP cells were from AntiCancer. MeWo cells were from CLS, and E0771 were from tebu-bio. B16-F10 and LLC1 cells were from ATCC. All tumor cell lines were cultured in DMEM supplemented with 10% FBS, penicillin/streptomycin (100 units/ml), and glutamine (2 mM). Cells were tested negative for mycoplasma contamination before experiments.

siRNA-mediated knockdown
Cells at 70% confluence were transfected with siRNAs using Opti-MEM (Thermo Fisher Scientific) and lipofectamine RNAiMAX (Invitrogen) according to the manufacturer's instructions. SiRNAs used for the screen are described in Table S1. SiRNAs directed against human CALCRL (EHU003121), human CCR2 (EHU109141), human GPR125 (EHU085311), human FZD6 (EHU074771), human determined by staining with an anti-Ki67 antibody (a), or the tumor cell growth was determined by measuring protein levels of GFP by Western blot analysis (b). The bar diagrams show the statistical evaluation (a; n = 3 independent experiments) or the relative densitometric values based on Fiji software (b; n = 3 independent experiments). (c and d) MLECs transfected with control siRNA or siRNA directed against Adm were cultured alone or together with B16-F10 cells (c) or LLC1 cells (d) transduced with scrambled control shRNA or shRNA directed against adrenomedullin. Thereafter adrenomedullin levels were determined by immunoblotting (c), or adrenomedullin concentrations in cell supernatant ware determined by ELISA (d). The bar diagrams in panel c show the relative densitometric values based on Fiji software (n = 3 independent experiments). (e) B16-F10 cells transduced with scrambled control shRNA (control B16) or shRNA directed against adrenomedullin (Adm-KD-B16) were injected subcutaneously in control or EC-Adm-KO mice, and tumor growth was determined (n = 6 mice for each condition). (f) LLC1 cells transduced with control shRNA (control LLC1) or shRNA directed against Adm (Adm KD LLC1) were injected subcutaneously in wild-type mice, and tumor growth was determined (n = 5 mice for each genotype). (g) Sections of tumors from control animals and EC-Adm-KO mice injected with control B16-F10 or Adm-KD-B16 cells were analyzed for markers of endothelial cells (PECAM1; green), perivascular cells (α-SMA; red) or for proliferating cells (Ki67; red), and were stained with DAPI (blue). The bar diagram shows the statistical analysis of the vessel area (f, n = 6 per mice group). Bar length: 100 μm (a), 50 μm (g). Data represent mean values ± SEM; *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001 (two-way ANOVA and Bonferroni's post hoc test [b and e-g], one-way ANOVA and Tukey's post hoc test [c and d], and two-tailed Student's t test [a]).

Nakayama et al.
Journal of Experimental Medicine

Co-culture experiments
For Ki67 staining or TUNEL staining, 1.5 × 10 3 GFP-expressing tumor cells were added onto the endothelial monolayer and cultured overnight. For Ki67 staining, cells were fixed for 10 min in 4% paraformaldehyde (PFA). After washing with PBS, cells were blocked and permeabilized in blocking buffer (0.3% Triton X-100 and 1% BSA in 1× PBS) at room temperature for 15 min. Thereafter, cells were incubated with primary antibody directed against Ki67 (Abcam) overnight at 4°C (dilution 1:100). After gentle washing with PBS (three times), cells were incubated with corresponding Alexa Fluor 594-conjugated secondary antibody (1:500; Invitrogen) together with DAPI (1 ng/ml; Invitrogen) for 1 h at room temperature.
For determination of GFP by immunofluorescence or Western blot analysis, and determination of ADM or CCL2 protein concentrations in cell supernatants, 1 × 10 5 endothelial cells and 1 × 10 5 tumor cells were cultured in the same well. For determination of GFP by immunofluorescence, cells were fixed for 10 min in 4% PFA and washed with PBS. After permeabilization in blocking buffer for 30 min as described above, cells were stained with DAPI for 1 h at room temperature.
TUNEL assay 1.5 × 10 3 GFP-expressing tumor cells were added onto the endothelial monolayer and cultured overnight. Cells were then fixed for 15 min in 4% PFA at room temperature. Apoptotic cells were detected using the Click-iT TUNEL Alexa Fluor 594 Imaging Assay (C10246; Invitrogen) following the manufacturer's instructions.

Determination of ADM and CCL2 level
Supernatants collected from endothelial cells and/or tumor cells were transferred to precooled tubes, and cellular debris was removed by centrifugation (20,000 g) for 10 min at 4°C. Adrenomedullin and CCL2 concentrations were determined with an adrenomedullin ELISA Kit (catalog #LS-F6083-1; BIOZOL Diagnostica Vertrieb GmbH) or an MCP-1/CCL2 Human Uncoated ELISA Kit (88-7399-22; Invitrogen) following the manufacturer's instructions.

Lentiviral infection of tumor cells
To generate tumor cell lines, in which expression of Adm or Ccr2 was suppressed, the lentiviral pLV-Puro-U6 expression vector containing the sequence of scramble shRNA, murine Adm shRNA or murine Ccr2 shRNA (VectorBuilder) were used. To generate EGFP-expressing tumor cell lines pLV-Bsd-EF1A expression vector containing the sequence of EGFP (VectorBuilder, VB210426-1127uue) was used.
Each vector was transfected into HEK293T cells along with envelope plasmid pMD2.G and packaging plasmid psPAX2. Following 48 h of incubation, the supernatant was collected and filtered through a 0.45-μm low-protein binding Durapore represents the relative densitometric values based on Fiji software (n = 3 independent experiments). (b) MLECs transfected with control siRNA (cont) or with siRNA directed against Gnas were cultured alone or together with LLC1 cells for 24 h, and ADM protein concentrations in cell supernatants was determined by ELISA (n = 5 independent experiments). (c and d) B16-F10 cells were cultured for 24 h with endothelial cell (EC) medium or with conditioned endothelial cell medium (CM) of MLECs transfected with control siRNA (CM control) or siRNA directed against Gnas (CM Gα s KD). ADM protein concentrations in cell supernatants were determined by ELISA (c; n = 5 independent experiments), or the expression level of ADM was determined by Western blot (d; n = 3 independent experiments). (e and f) HUVECs transfected with control siRNA or siRNA directed against GNAS or CALCRL were incubated without or with 10 nM of adrenomedullin for 3 h (e), or with 50 µM db-cAMP for 24 h (f). Thereafter, the CCL2 expression was determined by qRT-PCR analysis (n = 4 independent experiments). (g) HUVECs transfected with control siRNA or siRNA directed against GNAS or CALCRL or CCL2 were incubated without or with 10 nM adrenomedullin for 24 h, and the CCL2 concentration in cell supernatants was determined (n = 3 independent experiments). (h) Endothelial cells were isolated from B16-F10 tumors grown in control or EC-Gα s -KO mice by FACS, and expression of the indicated genes was determined by qRT-PCR (n = 3 mice for each genotype). Data represent mean values ± SEM; *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001; n.s., non-significant (two-way ANOVA and Bonferroni's post hoc test [a and e-g], one-way ANOVA and Tukey's post hoc test [b-d], and two-tailed Student's t test [h]). membrane (Millex). Tumor cells were transduced for 24 h with the lentiviral vector, followed by a complete medium change. After 48 h, cells were used for further analyses.
RNA sequencing RNA was isolated from ECs using the miRNeasy Micro Kit (Qiagen). Samples were treated with on-column DNase digestion (DNase-Free DNase Set, Qiagen), and total RNA and library integrity were verified with LabChip Gx Touch 24 (PerkinElmer). 2 μg of total RNA were used as input for TruSeq Stranded mRNA Library preparation following the Low Sample protocol (Illumina). Sequencing was performed on a NextSeq500 instrument (Illumina) using v2 chemistry. The resulting raw reads were assessed for quality, adapter content, and duplication rates. Only reads between 30 and 150 nucleotides were cleared for further analyses. Trimmed and filtered reads were aligned versus the Ensembl human genome version hg38 (GRCh38).

Quantitative real-time PCR analysis
Total RNA was isolated using the Quick-RNA Micro prep kit (Zymo) according to the manufacturer's protocol. Quality control of samples was carried out using a Nanodrop ND-100 Spectrophotometer. Complementary DNA synthesis was performed using the ProtoScript II Reverse Transcription kit (M0368S; New England BioLabs). Quantitative real-time PCR was performed using primers designed with the online tool provided by Roche and the Light-Cycler 480 Probe Master System (Roche). Each reaction was run in duplicates, and relative gene expression levels were normalized to GAPDH. Relative expression was calculated using the ΔΔCt method. Primer sequences used are described in Table S2.

Experimental mice
All mice were backcrossed onto a C57BL/6N background, and experiments were performed with littermates as controls. Mice were housed under a 12-h light/12-h dark cycle, with free access to food and water and under specific pathogen-free conditions unless stated otherwise. The generation of inducible endothelium-specific Gα s -deficient mice (Tie2-CreERT2;Gnas flox/flox ), endothelium-specific CALCRL-and adrenomedullin-deficient mice (Tie2-CreERT2;Calcrl flox/flox or Tie2-CreERT2;Adm flox/flox , respectively) was described previously (Iring et al., 2019). Mice carrying a floxed Ccl2 allele were purchased from The Jackson Laboratory (stock no. 016849) and crossed with inducible endothelium-specific Cre transgenic mice (Tek-CreERT2; Korhonen et al., 2009). MMTV-PyMT mice were obtained from The Jackson Laboratory. Cre-mediated recombination in the adult mice was induced by i.p. injection of tamoxifen (T5648; Sigma-Aldrich) dissolved in Miglyol 812 (1 mg per mouse per day) on 5 consecutive days.

Syngeneic tumor models
For syngeneic tumor experiments, age (8-10 wk) and sexmatched mice were used. 2.5 × 10 5 tumor cells cultured at 75-80% confluency were suspended in 50 µl PBS, unless otherwise described, and were injected subcutaneously at the right flank. Mice were monitored every day, and tumor length (L) and width (w) were measured with a caliper. Tumor volume was calculated in the following way: 0.5 × (L × w 2 ). Mice were sacrificed after 15 d and tumors were extracted and processed for immunostaining, Western blotting, or cell sorting.
For mammary fat pad injection, mice were anesthetized by subcutaneously injecting xylazin/ketamine (16 mg/kg, 120 mg/kg body weight respectively), or by isoflurane. After a small incision was made between the fourth nipple and the midline with a scissor, a pocket was formed by inserting a cotton swab moistened with PBS. Then the mammary fat pad was exposed using tweezers. 30 µl of E0771 cell suspension (10 million cells/ ml) mixed with 20 µl cooled matrigel was injected into the mammary fat pad by holding the needle horizontally. Skin was sutured with absorbable 6.0 silk suture (CatGut). After surgery, animals were monitored until recovery in a chamber on a heating pad.

MMTV-PyMT mice
Mice were monitored every day for tumor initiation and growth. All palpable masses were measured using external calipers, and tumor mass was calculated as described above. Mice remained in the study until the mass reached a total volume of 1.7 cm 3 unless tumor ulceration or other complications occurred. Mice were also evaluated for their behaviors, grooming activities, body conditions and body weight and kept in the study until termination criteria were met. After reaching any of the previously described criteria, each mouse was euthanized with CO 2 and tissue samples were collected for subsequent analyses.

Analysis of retinal angiogenesis
Cre-mediated recombination was induced by i.p. injection of tamoxifen dissolved in Miglyol 812 (0.5 mg per mouse per day) on postnatal day 1 to 3. On day 7 or 27, whole eyes were fixed with 4% PFA for overnight at 4°C. After washing in PBS, retinae were dissected, cut into four quadrants and blocked/permeabilized (1% BSA, 0.3% Triton, PBS) overnight at 4°C. On the following day, retinae were washed two times in Pblec buffer (1% Triton X-100, 1 mM CaCl 2 , 1 mM MgCl 2 , and 1 mM MnCl 2 in PBS [pH 6.8]) for 20 min and incubated with biotinylated determined by staining for Ki67 (a), and the number of GFP-expressing tumor cells was determined by immunofluorescence (b) or determined as the protein level of GFP by Western blot analysis (c). In addition, expression of ADM was analyzed by immunoblotting (d). The bar diagrams show the statistical evaluation (a and b; n = 3 independent experiments) or the relative densitometric values based on Fiji software (c and d; n = 3 independent experiments). (e) HUVECs were transfected with control siRNA or siRNA directed against GNAS and cells were co-cultured for 24 h with MeWo cells transfected with control siRNA or siRNA directed against CCR2. Thereafter, the expression level of ADM was determined by Western blot analysis. The bar diagrams represent the relative densitometric values of the band recognized by the anti-ADM antibody based on Fiji software (n = 3 independent experiments). Bar length: 100 μm (a and b). Data represent mean values ± SEM; *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001 (two-way ANOVA and Bonferroni's post hoc test [c and e], and one-way ANOVA and Tukey's post hoc test [a, b, and d]). isolectin-B4 (1:50; VectorLabs) for 2 h. Following five washes in blocking solution, retinae were incubated with Alexa Fluor streptavidin-conjugated antibodies (1:100; Molecular Probes) for 2 h, washed three times in blocking solution and were flatmounted on microscope glass slides with Fluoromount-G (0100-01; SouthernBiotech). All images shown are representative of the retinal vascular phenotype observed in at least four individual pups. All quantifications were done with Fiji software on high-resolution confocal images or lower resolution stereomicroscope images (vascular outgrowth). The radial length was defined as the distance from the center of the retina to the angiogenic front for each retina quadrant. The vascular density was defined as the ratio of isolectin B4 + area/total area.

Cell sorting and flow cytometric analysis
Mouse lung or tumors were dissected and enzymatically digested for 30 min while being shaken at 37°C in a digestion mix containing 0.25% collagenase II (Gibco) and 1 U/ml dispase (Gibco) in PBS. Cell suspensions were filtered through a 100-μm cell strainer followed by washing with 1% FCS in PBS. Antibodies used were as follows: Biolegend). Cells were sorted on an S3e cell sorter (BioRad) or FACSCanto (BD).

Study approval
All procedures of animal care and use in this study were approved by the local animal welfare authority committee of the Regierungspräsidium Darmstadt.

Statistics
Trial experiments or experiments done previously were used to determine sample size with adequate statistical power. Samples were excluded in cases where RNA/cDNA quality or tissue quality after processing was poor (below commonly accepted standards). Animals were excluded from experiments if they showed any signs of sickness. The investigator was blinded to the group allocation and during the experiment. Data represent biological replicates. In all studies, comparison of mean values was conducted with unpaired, two-tailed Student's t test or oneway or two-way ANOVA with Bonferroni's post hoc test. In all analyses, statistical significance was determined at the 5% level (P < 0.05). Depicted are mean values ± SEM as indicated in the figure legends. Statistical analysis was performed with Prism5 or Prism6 (GraphPad) or Excel (Microscoft) software. Fig. S1 shows the effects of endothelial loss of Gα s on retinal angiogenesis, tumor angiogenesis, and tumor microenvironments. Fig. S2 shows role of endothelial Gα s or CALCRL, both in vitro and in vivo. Fig. S3 shows the effect of endothelial-or tumor cell-derived adrenomedullin on endothelial or tumor cell proliferation. Fig. S4 shows tumor-suppressing function of endothelial CCL2 whose formation is inhibited through adrenomedullin signaling. Table S1 shows target sequences of siRNAs used in the siRNA screen in Fig. 3 a and Fig. S2 b. Table S2 shows primer sequences used for qRT-PCR. and Digestive and Kidney Diseases, National Institutes of Health. Open Access funding provided by the Max Planck Society.

Online supplemental material
Author contributions: A. Nakayama initiated the study, performed most of the in vitro and in vivo experiments, analyzed the data, and wrote the manuscript; K.A. Roquid, A. Iring, and B. Strilic helped with in vitro and in vivo experiments. S. Günther performed RNA sequencing and analyzed and discussed data; L.S. Weinstein and M. Chen provided tools and discussed data. S. Offermanns supervised the study, discussed the data and wrote the manuscript. All authors commented on the manuscript. Figure S3. Role of adrenomedullin in proliferation of endothelial and tumor cells. (a and b) HUVECs transfected with control siRNA or siRNA directed against ADM were cultured with GFP-MDA 231 for the indicated time periods, then either stained with anti-Ki67 antibody (red) and with DAPI (blue; a) or lysed to analyze GFP expression as an indicator of tumor cell number (b; n = 3 independent experiments). (c) HUVECs were co-cultured for 24 h with GFP-expressing MDA-MB-231 cells transfected with control siRNA or siRNA directed against ADM. Endothelial proliferation was determined by analyzing immunofluorescence obtained with the anti-Ki67 antibody. Bar diagrams show the statistical evaluation (n = 3 independent experiments). (d) HUVECs transfected with control siRNA or siRNA directed against Adm were cultured alone or together with MeWo cells transfected with control siRNA or siRNA directed against adrenomedullin. Thereafter, adrenomedullin levels were determined by Western blot analysis. (e) MLECs were transfected with control siRNA or siRNA directed against Gnas or Calcrl, then cultured without (basal) or with conditioned medium (CM) of B16-F10 cells transfected with control shRNA (control B16) or shRNA directed against Adm (Adm-KD-B16). Endothelial proliferation was determined by staining with anti-Ki67 antibody (n = 2 independent experiments). (f) Expression of ADM in control B16-F10 melanoma cells and in Adm-KD-B16 was determined by Western blotting. (g) Whole mount retinae of control and EC-Adm-KO mice at P7 stained with Isolectin B4. The bar diagram shows the quantification of vasculature parameters (n = 4 for each genotype). (h) Control B16-F10 or Adm-KD-B16 cells were cultured for 72 h, and cells were counted. The bar diagram shows the statistical evaluation (n = 2 independent experiments). Bar length: 100 μm (a, c, e, and g); and in panel h: 50 μm. Data represent mean values ± SD; *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001 (one-way ANOVA and Tukey's post hoc test [e], two-way ANOVA and Bonferroni's post hoc test [b and h] and two-tailed Student's t test [a, c, and g]).