SIRT1‐mediated deacetylation of FOXO3a transcription factor supports pro‐angiogenic activity of interferon‐deficient tumor‐associated neutrophils

Angiogenesis plays an important role during tumor growth and metastasis. We could previously show that Type I interferon (IFN)‐deficient tumor‐associated neutrophils (TANs) show strong pro‐angiogenic activity, and stimulate tumor angiogenesis and growth. However, the exact mechanism responsible for their pro‐angiogenic shift is not clear. Here, we set out to delineate the molecular mechanism and factors regulating pro‐angiogenic properties of neutrophils in the context of Type I IFN availability. We demonstrate that neutrophils from IFN‐deficient (Ifnar1−/−) mice efficiently release pro‐angiogenic factors, such as VEGF, MMP9 or BV8, and thus significantly support the vascular normalization of tumors by increasing the maturation of perivascular cells. Mechanistically, we could show here that the expression of pro‐angiogenic factors in neutrophils is controlled by the transcription factor forkhead box protein O3a (FOXO3a), which activity depends on its post‐translational modifications, such as deacetylation or phosphorylation. In TANs isolated from Ifnar1−/− mice, we observe significantly elevated SIRT1, resulting in SIRT1‐mediated deacetylation of FOXO3a, its nuclear retention and activation. Activated FOXO3a supports in turn the transcription of pro‐angiogenic genes in TANs. In the absence of SIRT1, or after its inhibition in neutrophils, elevated kinase MEK/ERK and PI3K/AKT activity is observed, leading to FOXO3a phosphorylation, cytoplasmic transfer and inactivation. In summary, we have found that FOXO3a is a key transcription factor controlling the angiogenic switch of neutrophils. Post‐translational FOXO3a modifications regulate its transcriptional activity and, as a result, the expression of pro‐angiogenic factors supporting development of vascular network in growing tumors. Therefore, targeting FOXO3a activity could provide a novel strategy of antiangiogenic targeted therapy for cancer.

Therefore, targeting FOXO3a activity could provide a novel strategy of antiangiogenic targeted therapy for cancer.
Angiogenesis plays an important role in tumor growth and metastasis. Tumor-associated neutrophils (TANs) are known to affect tumor angiogenesis, but how? In this study, the authors found that a lack of interferon-1 (IFN-1) activates a molecular cascade within TANs that involves Sirtuin 1 (SIRT1) and the transcription factor FOXO3a. This cascade, in turn, can stimulate the release of pro-angiogenic factors into the tumor microenvironment. These insights into the molecular mechanisms of TAN regulation suggest new therapeutic targets for anti-angiogenic cancer therapies.

| INTRODUCTION
All solid tumors, like normal organs, need a proper vasculature to grow beyond a limited size. 1 The initiation of tumor neovascularization involves the sprouting of pre-existent vascular endothelial cells and the recruitment of pericytes. 2 Lately, tumor-associated neutrophils (TANs) gain increasing attention as cells influencing tumor angiogenesis. In fact, these cells orchestrate a broad spectrum of anti-or protumor activities, depending on tumor environment. 3 In this context, we could previously demonstrate that Type I interferons (IFNs) impair pro-angiogenic activity of neutrophils. In agreement, IFN-deficient mice show elevated intratumoral accumulation of pro-angiogenic neutrophils that support tumor vascularization and growth. 4,5 Proangiogenic TANs show deregulation of multiple signaling pathways, including elevated Granulocyte colony-stimulating factor (G-CSF) signaling. 6 Many factors involved in this pathway demonstrate regulatory activity, such as Sirtuin 1 (SIRT1), which is a potent protein deacetylase, and was shown to regulate multiple cellular functions, such as survival, differentiation and glucose metabolism, 7 and to be overexpressed in solid tumors and hematopoietic malignancies. 8,9 Nevertheless, the exact mechanism responsible for pro-angiogenic bias of IFN-deficient neutrophils is not clear.
Considering the emerging importance of TANs in the regulation of tumor angiogenesis, we seek to delineate the underlying mechanism and factors supporting the angiogenic potential of these cells.
We could demonstrate here that dynamics of vessel development and pericyte maturation in melanoma depends strongly on IFN availability, and that tumor vasculature is established significantly earlier in Ifnar1 À/À animals. Moreover, we observe that pro-angiogenic Ifnar1 À/À TANs strongly support these processes via the release of angiogenic factors, and that their angiogenic potential is supported by the SIRT1-mediated activation of forkhead box protein O3a Littermates (8-12 weeks old) of C57BL/6J wild-type (WT) and Ifnar1 À/À strains were used in all experiments. Mice were bred and kept under specific pathogen free conditions in the animal facility of the University Hospital Essen (Essen, Germany). Experiments were done in accordance with government and institute guidelines and regulations.

| Immunohistochemistry and vascular network analysis
Tumors were dissected and snap frozen in liquid nitrogen. Consecutive cryosections (5 μm thickness) of murine tumors were fixed and processed for immunostaining as described before. 6  and analyzed with FIJI (ImageJ) 11 and AngioTool softwares. 12

| Flow cytometry
Single-cell suspensions from tumors were prepared; erythrocytes were removed using erythrocyte lysis buffer. Single-cell suspensions were prepared for flow cytometry staining (gating strategy Figure S1A,B).
Single-cell suspensions were stained in PBS with conjugated antibodies listed below, and flow cytometry was performed using BD

| TAN isolation
Heparinized blood was collected via heart puncture and tumors were harvested and digested using dispase II 0.  Figure S1).

| Immunofluorescence and subcellular localization
Isolated WT and Ifnar1 À/À tumor associated neutrophils, 15 000/well were incubated with or without treatment with 1 μm EX-527(SIRT1 Membranes were washed and HRP-conjugated antibodies added, allowing the detection of cytokines by enhanced chemiluminescence reaction. Quantification of signal intensity was done using FIJI (ImageJ) software. Signal intensities were normalized to the control spots intensities on the corresponding membranes.

| Aortic ring assay
Aortic ring assay was performed as described previously. 13 Shortly, thoracic aortae from WT mice were dissected, fat and connective tissue were removed and 0.5 mm aorta rings were placed in 96-well plates coated with type I collagen (Enzo life sciences, Farmingdale, NY). Once embedded, rings were covered with the endothelial cell growth medium (Promocell GmbH, Heidelberg, Germany). Subsequently, 2 Â 10 4 TANs were added (WT, Ifnar1 À/À or EX-527-treated Ifnar1 À/À ). The wells imaged 12 days after incubation in 37 C using AMG EVOS digital phase contrast inverted microscope and analyzed by FIJI (ImageJ) software.

| ELISA (Enzyme-linked Immuno Sorbent Assay)
TANs were isolated from WT and Ifnar1 À/À melanoma B16F10-bearing mice (as described above), and coincubated with B16F10 tumor cells in ratio 10 000:10 000 in 100 μL DMEMc for 24 hours, 10 000 cells B16F10 alone in 100 μL DMEMc were used as a control. Supernatants were collected and the concentration of Vegf and Mmp9 were estimated according to the manufacturer's protocol (R&D, Minneapolis, MN).

| Statistics
Statistical analyses were performed using Kruskal-Wallis ANOVA (englisch analysis of variance) method for multiple comparisons with the Bonferroni correction, Mann-Whitney U test was performed for two independent samples. P < 0.05 was considered significant.

| Type I IFN-deficiency drives vessel normalization in growing tumors
Type I IFN-deficient (Ifnar1 À/À ) mice show enhanced tumor growth after transplantation of B16F10 melanoma cells, when compared to their WT counterparts ( Figure 1A,B). This phenomenon is associated with an elevated infiltration of CD11b + Ly6G + neutrophils into the tumor site of Ifnar1 À/À mice ( Figure 1C,D), which accumulate preferentially in close vicinity to the tumor vascular network ( Figure 1E).
As tumor growth depends on the efficient vasculature, and this seems to be supported by pro-angiogenic action of neutrophils, we have assessed the vascular tumor network in Ifnar1 À/À vs WT mice. Interestingly,  whereas the number of vessels does not differ between WT and Ifnar1 À/À tumors (Figure 2A), the area covered by vascular network is significantly higher in tumors growing in Ifnar1 À/À mice ( Figure 2B). This was due to the increase in vessel perimeter in such animals ( Figure 2C). Thus, we next addressed the vascular function in both groups.
Stabilization and maturation of the vascular network depends on the recruitment of mural cells, so called pericytes, and vascular smooth muscle cells, to nascent blood vessels. 14 Therefore, we performed morphometric analyses by using the well-established pericyte markers, known to be expressed at different stages of vascular maturation: PDGFRb (Platelet-derived growth factor receptor beta, early stage), desmin (intermediate stage) and αSMA (smooth muscle actin, late stage). 15 In line with the enhanced vascular lumen in Ifnar1 À/À mice, we observed elevated vascular normalization in these animals, reflected by a significantly higher coverage by PDGFRb + and Desmin + pericytes, and enhanced αSMA + perivascular cells ( Figure 2D-I).

| Elevated pro-angiogenic protein expression in tumors growing in Ifnar1 À/À mice
Our data show that tumor angiogenesis is regulated by neutrophils. 4,5 As these neutrophils efficiently infiltrate tumor tissue, we evaluated the expression of pro-angiogenic proteins in the tumor tissue of WT and Ifnar1 À/À mice, at early (Day 10 postinjection) and late (Day 14) tumor development stage, using a Proteome Profiler Assay. In total, we detected differences in the expression of 38 proteins with angiogenic activity in tumor tissue ( Figure S3A). At Day 10, we could identify the upregulated expression of CCL22, CCL6, FGF-21 and VEGF in tumors from Ifnar1 À/À mice ( Figure S3B). These factors facilitate tumor progression, metastasis, [16][17][18] and promote angiogenesis. 19 At late stage of tumor development (Day 14), we detected in Ifnar1 À/À tumors upregulated Fetuin A, promoting tumor metastasis, 20 Angiopoietin-2 that modulates angiogenesis in cooperation with VEGF, 21 and tumor-promoting FGF-21 and IL-22 22 ( Figure S3B). This confirms higher angiogenic status of tumors growing in Ifnar1 À/À mice.

| Pro-angiogenic Ifnar1 À/À neutrophils show elevated activity of FOXO3a
Comparing of gene and protein expression in tumor lysates of WT and Ifnar1 À/À mice revealed significantly elevated expression of transcription factor FOXO3a in samples from Ifnar1 À/À animals ( Figure 3A). FOXO3a is involved in multiple cellular processes, including angiogenesis, 23 and its activity depends on its cellular localization and posttranslational modification. Phosphorylation of this factor by kinases leads to its inactivation and cytoplasmic retention, 24 while deacetylation, leads to FOXO3a activation and nuclear localization. 25 As tumors from Ifnar1 À/À mice show elevated level of FOXO3a, we assessed its activity and posttranslational modification in Ifnar1 À/À TANs, and compared it to WT counterparts. We have isolated TANs from tumors of Ifnar1 À/À or WT mice, and stain their FOXO3 and pFOXO3.
We could demonstrate significantly higher expression of active nuclear form of FOXO3a in pro-angiogenic Ifnar1 À/À neutrophils ( Figure 3B,D), while expression of inactive pFOXO3a was higher in the cytosol of WT TANs ( Figure 3C,E). In agreement, the expression of pFOXO3a and the ratio of inactive pFOXO3a to active FOXO3a, as assessed by flow cytometry, was significantly elevated in WT TANs ( Figure 3F,G).
In line with these findings, MEK/ERK and PI3K/AKT kinases, which are involved in FOXO3a phosphorylation, 26 were upregulated in WT tumor lysates, as compared to Ifnar1 À/À (Figure 4A,B). Thus, elevated kinase activity results in phosphorylation of FOXO3a and its repressed activity in WT neutrophils.

| Inhibition of FOXO3a deacetylation in pro-angiogenic Ifnar1 À/À neutrophils leads to its cytosolic transfer and inactivation
Deacetylation of FOXO3a is mediated by sirtuins, such as SIRT1, and leads to its activation and nucleus retention. 27,28 In agreement with this, Ifnar1 À/À TANs, which are characterized with higher level of active nuclear FOXO3a, show a significant upregulation of SIRT1 deacetylase, as compared to WT ( Figure 5A).
To proof the role of SIRT1 in the activation of FOXO3a in TANs, we inhibited this factor using its molecular inhibitor EX527 and evaluated the expression and activity of FOXO3a. Indeed, we could observe diminished FOXO3a gene expression and its reduced nuclear retention in treated Ifnar1 À/À cells ( Figure 5B-D). Moreover, elevated levels of inactive pFOXO3a were detected in the cytosol of EX527-treated neutrophils, as compared to untreated cells ( Figure 5E,F).
F I G U R E 5 Legend on next page.
In agreement with our hypothesis that active FOXO3a supports angiogenic properties of TANs, treatment with SIRT1 inhibitor decreased the expression of pro-angiogenic factors in such TANs. The expression of VEGF and MMP9 in Ifnar1 À/À TANs treated with EX527 was comparable to WT TANs ( Figure 5G).
To evaluate the effect of EX527-treatment on pro-angiogenic functionality of TANs, we performed quantitative aortic ring angiogenesis assay using TANs as angiogenic stimulus. This assay allows ex vivo assessment of all steps of efficient angiogenesis, including endothelial cell proliferation, migration, tube formation, microvessel branching, perivascular recruitment and remodeling. 13 TANs from WT and Ifnar1 À/À mice were isolated, treated with EX527 and applied on aortic rings. After 14 days of coculture, number and length of developed microvessels were evaluated. We could observe significantly elevated angiogenic activity of Ifnar1 À/À TANs in comparison to WT TANs ( Figure 5H,I). Moreover, treatment of Ifnar1 À/À TANs with EX527 significantly reduced the proangiogenic capacity of such TANs, which was now comparable to WT levels. As expected, inhibition of SIRT1 in WT TANs has shown no visible effect on angiogenic properties of TANs, possibly due to already low amounts of this molecule in WT neutrophils.

| TANs with inactive FOXO3a fail to support tumor angiogenesis and growth when transferred into tumor-bearing mice
To confirm the role of SIRT1-mediated activation of FOXO3a in pro-angiogenic capacity of neutrophils, a rescue experiment was performed. Neutrophils were isolated from WT or Ifnar1 À/À mice, incubated with EX527 as described previously, mixed with B16F10 cells and injected s.c. into WT mice. At Day 3 p.t., tumor-bearing mice received EX527-treated or medium-treated WT or Ifnar1 À/À neutrophils (i.v.), respectively. Tumor growth was monitored for 14 days, mice sacrificed, tumors removed and tumor size assessed. We observed that the treatment of mice with Ifnar1 À/À TANs significantly accelerates tumor growth, while the pretreatment of Ifnar1 À/À TANs with EX527 prior to injection into mice, abolished this effect ( Figure 6A,B). As expected, mice injected with WT neutrophils show decreased tumor growth and EX527-treatment of WT TANs did not significantly change their effect on tumor growth. Therefore, we have excluded this group (EX527-treated WT) from further analyses.
In line with our ex vivo data, mice injected with EX527-treated Ifnar1 À/À neutrophils showed significantly reduced maturation of tumor vasculature, in comparison to mice injected with untreated Ifnar1 À/À neutrophils ( Figure 6C-K), indicating once more that SIRT1-mediated deacetylation is required for the activation of FOXO3a and pro-angiogenic function of TANs.

| Regulation of angiogenic properties of TANs by post-translational modification of FOXO3a is a common mechanism in preclinical melanoma
We next asked if our observation in B16F10 melanoma model is cell line-or tumor entity-specific. Therefore, we extended our analyses by using a second syngeneic melanoma model-CM. CM melanoma cells were s.c. injected into Ifnar1 À/À and WT control mice, as described for B16F10, and tumor growth was monitored for 14 days ( Figure S4A). After 14 days, mice were sacrificed, tumors removed, tumor weight and tumor angiogenesis assessed. CM tumors growing in Ifnar1 À/À mice showed significantly accelerated growth in comparison to WT mice, similar to B16F10 cells. Tumors in Ifnar1 À/À mice grew faster and reached significantly larger sizes.
In addition, higher numbers of TANs infiltrated such tumors ( Figure S4B). In line, immunohistological and morphometric analyses showed more advanced tumor angiogenesis in Ifnar1 À/À mice ( Figure S4C-E), as predicted.
In agreement with B16F10 data, analysis of FOXO3a activation in TANs showed a significantly enhanced nuclear retention of active FOXO3a in Ifnar1 À/À neutrophils ( Figure S5A,B), while WT cells exhibit high levels of inactive cytosolic pFOXO3a ( Figure S5C,D).
Thus, regulation of angiogenic properties of neutrophils by posttranslational modification of FOXO3a appears to be a general phenomenon in preclinical melanoma models.
F I G U R E 5 Inhibition of SIRT1 activity by EX527 leads to elevated phosphorylation of FOXO3a in Ifnar1 À/À TANs and inhibits their proangiogenic activity. (A) Elevated expression of SIRT1 in Ifnar1 À/À TANs. (B) FOXO3a expression in Ifnar1 À/À TANs is downregulated by SIRT1 inhibition. TANs from WT and Ifnar1 À/À mice were sorted. Half of Ifnar1 À/À TANs were treated in vitro with SIRT1 inhibitor EX527. Relative expression of FOXO3a gene expression was assessed by real-time quantitative PCR (RT qPCR). (C, D) Elevated nuclear FOXO3a intensity in Ifnar1 À/À TANs. Isolated TANs were stained for FOXO3a protein and nuclear FOXO3a intensity was quantified. Representative immunofluorescence images (C) and quantification (D) are shown. (E, F) FOXO3a phosphorylation and cytosolic localization is elevated in TANs after inhibition of SIRT1 activity. Representative immunofluorescence images (E) and quantification of cytosolic pFOXO3a intensity (F) in Ifnar1 À/À and Ifnar1 À/À TANs treated with EX527. Scale bars: 2 μm. For comparison of multiple and two independent groups, Kruskal-Wallis test and Mann-Whitney U test were used, respectively. Data are shown as mean ± SEM, and median, interquartile range, minimal and maximal values, *P < .05, ***P < .001. (G-I) SIRT1 inhibition suppresses angiogenic capacity of TANs. WT and Ifnar1 À/À TANs were isolated, treated with EX527, and their pro-angiogenic gene expression assessed (G) WT TANs, WT TANs treated with EX527, Ifnar1 À/À TANs or Ifnar1 À/À TANs treated with EX527 were incubated with aortic rings as described in Section 2. Quantification (H) of number and length of endothelial sprouts after incubation and representative images (I). Scale bars: 200 μm. For comparison of multiple independent groups, Kruskal-Wallis test was used. Data are shown as mean ± SEM, and median, interquartile range, minimal and maximal values, *P < .05; **P < .01. TANs, tumor-associated neutrophils; WT, wild-type Sirt1 expression, suggesting similar regulatory mechanism in human.
In addition, a significant negative correlation between Foxo3 and type I IFNs is observed in melanoma ( Figure S6A,B). Interestingly, this phenomenon seems not to be limited to melanoma cancer, as samples obtained from HNC (head-and-neck cancer) patients show comparable correlation between Foxo3 and Sirt1, and Foxo3 and Ifnar1 ( Figure S6C,D). Data for HNC were obtained from the available Gene Expression Omnibus database GSE65858 (reference Figure S6).

| DISCUSSION
Tumor neovascularization is a complex process relying on the prolifer- show that IFN-deficient TANs efficiently support tumor angiogenesis and that removal of such cells suppresses this process. 4 We could also show that tumor-associated macrophages do not play a significant role in our experimental system, as the depletion of neutrophils aborted angiogenesis and tumor growth, clearly showing no involvement of macrophages. 6 Moreover, transfer of pro-angiogenic Ifnar1 À/À neutrophils into WT tumor bearing mice significantly elevates tumor angiogenesis in such mice. 6 However, the mechanism controlling angiogenic activity of TANs has not yet been identified.
Therefore, we focused here on the morphometric vascular situation in the context of intratumoral neutrophils. The plasticity window for vascular remodeling in tumors is defined by the coverage of pericytes.
The extent of such coverage is an important marker of vessel maturation 30 and its absence is responsible for the irregular, tortuous and leaky blood vessels in tumors, modulating therapy efficacy. 31 Several molecules have been proposed as pericyte markers during maturation of the vascular bed. 32 NG-2, Desmin and PDGFRβ have been established as markers of early, that is, immature pericytes, whereas αSMA has been reported as a marker of mature mural cells, including pericytes and smooth muscle cells. 30,32 In our study, using two different melanoma cell lines, we observed elevated neutrophil infiltration into tumor tissue in Ifnar1 À/À mice, compared to WT. Importantly, such neutrophils accumulated in close proximity to vessels and secreted high amounts of pro-angiogenic factors that support proliferation and sprouting of endothelial cells, but also enhance recruitment of pericytes to the vascular wall. In a consequence, a highly mature vascular network and normalization of the vascular bed is observed in such Ifnar1 À/À tumors.
In highly angiogenic tumors of Ifnar1 À/À mice, we have observed elevated hypoxia even though the vasculature was normalized. Possibly, the accelerated tumor growth that is observed from Day 9 p.t. could argue for this phenomenon. We hypothesize that due to the We could demonstrate a significant upregulation of proangiogenic genes encoding VEGF, MMP9, BV8 and S100a8 in Ifnar1 À/À TANs. MMP9 is known to degrade extracellular matrix and so to release sequestered growth factors, such as VEGF, to promote angiogenesis. 33 40 The impact of such modifications is diverse, and so kinase-mediated phosphorylation of FOXO3a results in its inactivation and cytoplasmic retention 24 of this factor, similarly acetylation via CBP/p300 attenuates FOXO3a transcriptional activity, while deacetylation activates this factor. 26,41 It has been reported that phosphorylation-mediated inactivation of FOXO3a occurs through the phosphatidylinositol-3-kinase/AKT (protein kinase B) and the RAS mitogen-activated protein kinase pathways. 42 Cytoplasmic retention and degradation of the phosphorylated FOXO3a inhibits transcription of target genes. 43 By contrast, oxidative or nutrient stress signals induce activation of JNK, MST1 and AMPK, which phosphorylate FOXO3a at the conserved residue sites, resulting in nucleus localization of FOXO3a and promoting the transcription of target genes.
Moreover, it has been shown that FOXO3a protein is degraded by AKT-triggered ubiquitination-proteasome pathway. 43 In agreement, we observed elevated phosphorylation of MEK/ERK and PI3K/AKT kinases in WT TANs. This leads to the phosphorylation of FOXO3a, its cytoplasmic translocation and inactivation. Its inactive form fails to activate pro-angiogenic genes, resulting in impaired angiogenic activity of neutrophils and suppressed tumor angiogenesis.
Our study shows a significant upregulation of SIRT1 in TANs of Ifnar1 À/À mice. This is accompanied by the elevated amount of active form of FOXO3a in these cells, and its diminished phosphorylation.
SIRT1 is a nicotinamide adenine dinucleotide (NAD + )-dependent protein deacetylase, which was shown to efficiently deacetylate FOXO3a, 27 resulting in its increased activity. 25 In contrast, acetylation of FOXO3a attenuates its transcriptional activity 44 and induces its proteasomal degradation. 45 It has been reported that SIRT1-mediated deacetylation activates FOXO3a in the nucleus, leading to the expression of several targets genes. 25 Moreover, increased SIRT1/ SIRT2-FOXO3a interaction was shown to tightly bind FOXO3a to DNA binding domain and so to enhance transcription of target genes. 26,41 Our findings are in line with these studies. In addition, we could observe strong positive correlation between SIRT1 and FOXO3 expression in human tumor tissue-both in melanoma, but also in HNC. Type I IFNs seem to regulate this axis also in human specimen, as significant negative regulation between FOXO3 and IFNs was observed.
While comparing angiogenic potential of TANs from WT vs Ifnar1 À/À mice, we observed significantly elevated capacity of Ifnar1 À/À neutrophils to stimulate endothelial cell proliferation and sprouting in aortic ring assay. Blocking the activity of SIRT1 in Ifnar1 À/À TANs led to the cytoplasmic translocation of FOXO3a, its elevated phosphorylation and deactivation. This was accompanied by the ceased expression of angiogenic mediators and suppressed angiogenic capacity of such cells.
Yet, the role of SIRT1-mediated deacetylation of FOXO3a is controver- Taking together, our findings demonstrate an essential role of FOXO3a transcription factor in the stimulation of pro-angiogenic functions of TANs. Activity and DNA binding affinity of FOXO3a is controlled by the fine balance between acetylation and phosphorylation. 41 Deactivation of FOXO3a could reprogram TANs into antiangiogenic phenotype and thus lead to the suppression of tumor angiogenesis and growth. This could offer new roads to therapeutically target tumor angiogenesis.