Pro-atherogenic diet accelerates tumor growth in mice through IL-1β and myeloid cell-derived VEGF-A

Aims: Myeloid inammatory cells are recruited to the tumor microenvironment and subsequently educated in situ to acquire a pro-invasive, pro-angiogenic and immunosuppressive phenotype. Components of the metabolic syndrome are known to aggravate tumorigenesis in part through myeloid cell activation. We hypothesized that consumption of a high fat/high cholesterol pro-atherogenic diet and its associated low-grade inammation would accelerate the initiation of solid tumors. Methods and results: Here, we show that two-week feeding of wildtype C57BL/6J mice with a proatherogenic diet increases the pool of circulating inammatory Ly-6C hi monocytes available for initial melanoma development and amplies the accumulation of myeloid cells within the tumor microenvironment, in an IL-1β-dependent manner. Under pro-atherogenic diet feeding, myeloid cells display heightened pro-angiogenic, pro-inammatory and immunosuppressive activities. Within the rst days after tumor implantation, myeloid cells become the main producer of VEGF-A in the tumor. Depletion of Ly-6C hi monocytes in mice fed with a pro-atherogenic diet limits immune cell inltration in the tumor, and inhibits tumor growth. IL-1β deciency or specic inhibition of VEGF-A in myeloid cells recapitulates the benecial effect of Ly-6C hi monocyte depletion, suggesting their complementary roles in tumorigenesis in the context of mild hyperlipidemia. Conclusion: Our study shows that dyslipidemia provide high amounts of activated myeloid cells with protumoral activity and shed light on cross-disease communication between cardiovascular pathologies and cancer. Translational Perspective: In this study we demonstrate that dyslipidemia accelerates the development of solid tumors through the increased inltration of Ly6C hi monocytes that differentiate into pro-tumoral myeloid cells. These ndings demonstrate that dyslipidemia can silently boost tumor development in normal-weight individuals through the action of IL-1β and VEGF-A. Our work sheds light on the potential benet of targeting IL-1β and VEGF-A in cancer patients with moderate dyslipidemia. Genetically-modied mouse models of atherosclerosis, such as apoe -/- or ldlr -/- present a strong hypercholesterolemia. The use of an atherogenic HFHCD aggravates hypercholesterolemia and hypercholesterolemia-associated monocytosis in these mice, which accelerates the evolution of atherosclerosis 23,24 . Here we found that feeding non-transgenic C57BL/6J mice with HFHCD for 2 weeks also led, though to a lesser extent, to increased plasma cholesterol levels (1.26±0.1g/l in HFHCD-fed mice (n=10) versus 0.56±0.13g/l in control chow diet (CD)-fed mice (n=13); P ≤ 0.0001) and monocytosis (6.8±1.6% in HFHCD-fed mice (n=8) versus 4.8±1% in control CD-fed mice (n=16); P ≤ 0.01). Ly6C hi monocytes dominated HFHCD-associated monocytosis over Ly6C lo monocytes, as seen by their increased percentage and number in blood (Figure 1A). Sustained elevation of circulating Ly6C hi monocytes has been identied in several pathologies as a marker of chronic inammation with a prognostic value 25 . Based on this observation, we aimed to evaluate the role of hyperlipidemia-associated systemic low-grade inammation on solid tumor development. We subcutaneously injected B16-F10 melanoma or TC1 tumor cells to mice under either CD or 2-week-HFHCD, and followed tumor growth for up to 15 days (Figure 1B). Mouse euthanasia was performed by cervical dislocation after anesthesia with 4% isourane. We found HFHCD signicantly accelerated


Introduction
Metabolic syndrome (MS) is an established risk both for cancer and cardiovascular disease (CVD), the 2 leading causes of morbidity and mortality worldwide. Components of the MS include excess body fat, increased blood pressure, high blood sugar and dyslipidemia 1 . The metabolically unhealthy phenotype is de ned by two or more of four of these components. In several types of cancers, dyslipidemia and obesity are associated with increased death rate 2,3 . This has also been functionally validated in experimental models of diet-induced obesity (DIO), which showed accelerated tumor growth 4,5 . Even though obesity is often associated with metabolic disorders, 24 to 30% of normal-weight adults are considered metabolically unhealthy, according to several worldwide meta-analyses 6,7 . These metabolically unhealthy normal weight individuals are commonly dyslipidemic. They present high circulating levels of cytokines and cardiometabolic risks comparable to those observed in metabolically unhealthy obese individuals 8 . This highlights the need for a better understanding of the impact of MS and its individual components in tumorigenesis in normal-weight individuals. Dysfunctional fatty acid and cholesterol metabolism promotes tumor growth directly through activation of oncogenic signaling pathways and formation of lipid rafts 9,10 , and indirectly through increased production and mobilization of monocytes 11,12 . Monocyte and monocyte-derived cells accumulate in many types of human and murine tumors and are thought to regulate several steps of tumor development, including antitumor T cell responses and angiogenic processes. Generally, high numbers/percentages of circulating monocytes indicate poor clinical prognosis in cancer 13 , suggesting that the level of monocytes in blood may in uence tumor development. In line with this hypothesis, it was recently shown that the peripheral count of monocytes was associated with the density of tumor-associated macrophages in colorectal cancer 14,15 . During in ammation, metabolic and epigenetic rewiring has been identi ed as important mechanisms of monocyte, and more generally myeloid cell, production and function 16 . Interleukin-1β (IL-1β), primarily generated by myeloid cells, functions as an upstream alarm capable of simultaneously regulating in ammation, mainly by inducing a local cytokine network and enhancing in ammatory cell in ltration to affected sites, and promoting angiogenesis, either directly or indirectly via induction of proangiogenic factors such as vascular endothelial growth factor A (VEGF-A) (reviewed in 17 ). VEGF-A is produced by many cell types, including tumor cells and myeloid cells. As a major inducer of blood vessel growth, VEGF-A is involved in tissue remodeling in the context of cancer and cardiovascular diseases. In endothelial cells, VEGF-A and IL-1β share common signaling pathways and may synergistically regulate the expression of in ammatory genes and growth factors 18 . Anti-angiogenic agents and immunotherapies targeting VEGF-A are thought to create a transient window of vessel normalization, which might facilitate the diffusion of therapeutic agents 19,20 . Anti-IL-1β therapies, initially developed in the context of autoimmune diseases, have recently gained a great interest in the cardiovascular and oncology elds 21 . The major randomized trial on the role of IL-1β inhibition in atherosclerosis, named CANTOS, reported a reduction in the number of incident cases of lung cancer in patients treated with the anti-IL-1β antibody compared to placebo-treated patients 22 . This clinical study proposed that IL-1 antagonism could prevent both the recurrence of cardiovascular events and tumor development in patients with a persistent pro-in ammatory response.
Having reviewed the state of the art, we thought to evaluate the impact of mild-dyslipidemia-associated in ammation on solid tumor growth in mice. To address this question, we employed a high fat/high cholesterol atherogenic diet (HFHCD), known to increase lipid levels and induce systemic low-grade in ammation.
We found that 2-weeks feeding with HFHCD accelerated solid B16 melanoma development in C57BL/6J mice and induced an angiogenic and in ammatory signature in the tumor. Acceleration of tumor growth was associated with increased Ly-6C hi monocyte levels in blood and predominant accumulation of myeloid-derived cells in the growing tumor. Myeloid cells signi cantly contributed to the increase of VEGF-A in the tumor under HFHCD. IL-1β de ciency markedly reduced monocyte level in blood and subsequently stopped the expansion of tumors under HFHCD. Speci c loss of myeloid-derived VEGF-A recapitulated the protective effect of systemic myeloid cell depletion in mice under HFHCD. This study shows that silent low-grade in ammation fuels early growing solid tumors with a pro-tumoral contingent of pre-educated myeloid cells via IL-1β production, and promote tumorigenesis through increased immunosuppression and VEGF-A synthesis.

Animals
All experiments were conducted according to European Community for experimental animal use guidelines (EC2010/63), and have been approved by the Ethical committee of University Paris Descartes (CEEA 34) and the French Ministry of Agriculture (agreement 17112). To investigate the effect of The B16F10 and TC1 cells were purchased from ATCC (CRL6475, CRL2493). Cells were cultured in RPMI supplemented with 10% fetal bovine serum (FBS), 2 mM glutamine, 50 U/ml penicillin and 100μg/ml streptomycin (RPMI complete medium). Mice were subcutaneously (s.c) injected with 0.25x10 6 B16-F10 or TC1 cells in 100µl of saline buffer in the shaved abdominal ank. Tumor growth was monitored every 2-3 days using a caliper until 15 days post-injection. Tumor size was calculated as: Width x Length in mm 2 .
Cells isolation, staining and ow cytometry Blood samples were collected by submandibular punction with 20-gauge needle under iso uorane anesthesia. Spleens were dissected and pressed through a 40-μM cell strainer, red blood cells were lysed with osmotic lysis buffer. Tumor were isolated from mice, minced and placed into GentleMACS C-tube with PBS-FCS 2%, dissociated mechanically with GentleMACS dissociator (Miltenyi) according to manufactor's standard protocol, then ltered on 70um strainer.
For Dimensionality reduction using t-SNE and automatic clustering : CD45 live cells were manually gated from multicolor ow cytometry and exported in a FCS le. t-Distributed Stochastic Neighbor Embedding (t-SNE) dimensionality reduction was performed using bh tsne, an implementation of t-SNE via Barnes-Hut approximations. 20 000 events were used for t-SNE dimensionality reduction.

Gene expression analysis
For the transcriptomic analysis, 4 tumors per group were harvested from mice fed with CD or HFHCD at Day 9 post injection of B16-F10 melanoma cells. Tumors were minced and lysed into RLT buffer containing 1% β-mercaptoethanol. RNA was extracted using Qiagen RNeasy kit according to the manufacturer's instructions.
Analysis of gene was done by affymetrix microarrays at de GENOM'IC core facility (Cochin Institute) In vivo cells labelling A bead labeling technique was used to label monocytes in vivo, by retro-orbital i.v. injection of 1 μm uorescent latex microbeads (YG, Polysciences) diluted ¹⁄ in sterile PBS, on day 8 post tumor injection. Mice were sacri ced 24h later.
For intravital neutrophil and monocyte labeling, mice were retro-orbitally injected i.v with anti-Ly6G PE and anti-CD115 APC (5ug each) on day 8 post tumors injection. 6h later mice were sacri ced. Labeled cells were analyzed in blood and tumors by ow cytometry.

Culture in vitro
Bone marrow cells were collected from femurs and tibias by insertion of a needle into the bone and ushing with RPMI supplemented with 0.2% BSA and 1% FCS as previously described. Macrophages were differentiated after culture for 9 days with 20% L929 conditioned medium.
For B16-F10 in vitro proliferation, 10 000 B16F10 cells were co-cultured with 2 500 CD45 cells isolated from tumors or with BM-DM, in RPMI complete medium. 24h later, B16F10 proliferation was measured by 3T-Thymidine incorporation or MTT assay (Sigma). For Thymidine incorporation one ìCi [ 3 H] thymidine was added to each well during the last 18 h. Thymidine incorporation was assessed using a TopCount NXT scintillation counter (Perkin Elmer).
Cytokines and chemokines assay, ELISA Total protein from tumors was extracted using Bio-plex lysis buffer (Bio-Rad) according to manufactor's instructions, and concentration was determinated with BCA assay (Pierce). CD45+ cells were isolated from fresh tumors by magnetic sort (Mojosort mouse CD45 nanobeads, Biolegend). Cells were cultured for 24h in RPMI complete medium then supernatent were harvested.

Lectin immunostaining
Tumors were frozen in OCT compound, then prepared in 10 µm sections on a cryostat microtome. Lectin staining was done with Griffonia Simplicifolia Lectin I (GSL I) isolectin B4 Fluorescein (Vector Laboratories) according to manufactor's protocol. Quanti cation was done with Image J.

Doppler imaging and quanti cation
Mice were shaved, anesthetized with iso urane and kept under monitored temperature during the time of imaging. Vessel density was evaluated by laser-Doppler perfusion imaging to assess in vivo tissue perfusion in the tumor (Moor Instrument). Quanti cation of the vessel density was performed with MoorLDIReview V6.1 software.
2'-deoxy-2'-[18F] uoro-D-glucose (FDG) positron emission tomography-computed tomography (PET-CT) imaging Prior to each imaging session, mice were fasted overnight with water ad libitum. Mice were anesthetized with iso urane (IsoVet 100%; Centravet, France) in 100% oxygen (4% iso urane for induction; 1-2% for maintenance). Mice were weighted, and placed on a heated plate (Minerve, France). Glycemia was measured in blood drawn from the caudal ventral artery using an Accu-Chek® Aviva Nano A (Accu-Chek, France). A customized catheter with a 29 G needle (Fischer Scienti c, France) connected to a 5-cm polyethylene tubing (Tygon Microbore Tubing, Cole-Parmer0.020" × 0.060"OD; Fisher Scienti c, France) was installed in the lateral tail vein of the mice. 10 MBq of FDG ([Gluscan, Advanced Applied Applications, France) in 200 µl of saline solution were injected in the mice. The mice were then put back in their cages and left awake for 30 min. Mice were anesthetized again and installed in prone position in the imaging bed of the camera (NanoScan PET-CT, Mediso Medical Imaging Systems, Hungary). Respiration and body temperature were registered. Body temperature was maintained at 37°C and anesthesia was controlled on the breathing rate throughout the entire examination. CT scans were performed rst using the following parameters: mode semi-circular, tension of 50 kV, 720 projections full scan, 300 ms per projection, binning 1:4. CT data were reconstructed using ltered back projection ( lter : Cosine ; Cutoff : 100%). List-mode PET data were collected between 45 and 60 min post injection of FDG, binned using a 5-ns time window, a 400-to 600-keV energy window, and a 1:5 coincidence mode. In vivo PET acquisitions were reconstructed using the Tera-Tomo reconstruction engine (3D-OSEM based manufactured customized algorithm, Mediso Medical Imaging Systems, Hungary) with expectation maximization iterations, scatter and attenuation corrections. Images were analyzed using the software PMOD (PMOD Technologies LLC, Switzerland). Standardized Volume of Interest (VOI) was drawn in each organ of interest and Standardized Uptake Values (SUV) were calculated by dividing the mean tissue radioactivity concentration by the whole body concentration of the injected radioactivity. The Peak value was calculated as the maximum average SUV within a 1-cm3 spherical volume of interest and the tumor volume was automatically segmented at 33% of this value. Total FDG uptake was estimated as the product from the volume by the mean uptake of the segmented region.

Statistics
Graphs were generated and statistical analysis were performed with Prism software (GraphPad software, La Jolla). Results are expressed as means ± SEM. The Mann-Whitney t-or ANOVA tests were used as speci ed in the legends. Comparison between tumor growth curves have been performed using a twoway ANOVA test, and multiple comparisons have been corrected with the Bonferroni coe cient. The association between two variables was done by spearman correlation. P values less than 0.05 was considered signi cant

Results
Consumption of an atherogenic diet causes low-grade systemic in ammation and accelerates murine B16-F10 melanoma tumor development in C57BL/6J mice Genetically-modi ed mouse models of atherosclerosis, such as apoe -/or ldlr -/present a strong hypercholesterolemia. The use of an atherogenic HFHCD aggravates hypercholesterolemia and hypercholesterolemia-associated monocytosis in these mice, which accelerates the evolution of atherosclerosis 23,24 . Here we found that feeding non-transgenic C57BL/6J mice with HFHCD for 2 weeks also led, though to a lesser extent, to increased plasma cholesterol levels (1.26±0.1g/l in HFHCD-fed mice (n=10) versus 0.56±0.13g/l in control chow diet (CD)-fed mice (n=13); P≤0.0001) and monocytosis (6.8±1.6% in HFHCD-fed mice (n=8) versus 4.8±1% in control CD-fed mice (n=16); P≤0.01). Ly6C hi monocytes dominated HFHCD-associated monocytosis over Ly6C lo monocytes, as seen by their increased percentage and number in blood ( Figure 1A). Sustained elevation of circulating Ly6C hi monocytes has been identi ed in several pathologies as a marker of chronic in ammation with a prognostic value 25 . Based on this observation, we aimed to evaluate the role of hyperlipidemiaassociated systemic low-grade in ammation on solid tumor development. We subcutaneously injected B16-F10 melanoma or TC1 tumor cells to mice under either CD or 2-week-HFHCD, and followed tumor growth for up to 15 days ( Figure 1B). Mouse euthanasia was performed by cervical dislocation after anesthesia with 4% iso urane.
We found that HFHCD signi cantly accelerated solid melanoma growth after 9 days of tumor development ( Figure 1C, supplemental g 1A), and increased tumor weight at day 15, when tumors reached about 200 µm 2 ( Figure 1D). Similar results were observed in TC1-injected mice (Supplemental g 1B), suggesting that the protumoral effect of the HCHFD may be applicable to other types of solid tumors. We chose to focus on the melanoma model for the rest of the study. The size of tumors positively correlated with the weight of spleens, supporting the involvement of in ammatory mechanisms in tumor growth acceleration under HFHCD ( Figure 1E). We performed a transcriptome analysis at day 9 to provide valuable information about changes in gene expression at the early stages of tumorigenesis in response to HFHCD. The most signi cant genes were sorted by log fold change (logFC), clustered in a hierarchical manner using the rlog differences and shown in a heatmap (Figure 2A). Notable modi cations in the expression of genes involved in complement pathway, myeloid and T cell migration and genes regulating in ammation in tumors from the HFHCD group, in comparison to the CD group were observed. Clusters of expression are detailed in supplementary tables 1 and 2. VEGF-A expression was signi cantly increased in tumors from the HFHCD group in comparison to the CD group ( gure 2B, supplemental table 2). Consistently, VEGF-A protein level was also increased in tumors from the HFHCD group ( Figure 2C).
VEGF-A is a key player in tumorigenesis 26 and its expression level is associated with the progression of malignant melanoma and other types of cancers 27,28 . In line with the increase of VEGF-A, microvessel density was higher in tumors of the HFHCD group (Supplemental g 1C), suggesting an ampli cation of the angiogenic processes in response to HFHCD. The production of VEGF-A by tumor cells is generally stimulated by hypoxia and is associated with increased glucose metabolism 29 . We then tested whether the HFHCD directly ampli ed tumor cell activity in vivo. To this end, we measured [ 18 F] uorodeoxyglucose (FDG) uptake by the tumor at day 9 of tumor development, using positron emission tomography-computed tomography (PET-CT). Total, but not mean intensity, FDG uptake was higher in tumors of HFHCD-fed mice compared to tumors of CD-fed mice, which was mostly attributable to the increase in tumor size rather than to a change in metabolic reprogramming of tumor cells (Supplemental Figure 1D). This suggests that the pro-atherogenic HFHCD did not directly impact the metabolic activity of tumor cells in vivo, and rather points to the implication of other components of the tumor microenvironment in tumor aggravation under HFHCD.

HFHCD-associated myelopoiesis accelerates the recruitment of protumoral myeloid cells into the tumor microenvironment
We studied the early chronology of leukocyte accumulation in the growing melanoma of mice on CD or HFHCD. Using ow cytometry, we found a progressive accumulation of CD45 + leukocytes within the growing tumors over time, showing that in ammatory cells continuously in ltrate the tumor ( Figure 3A). Nevertheless, leukocytes accumulated more extensively after 7 days in tumors of mice fed a HFHCD diet, suggesting a temporal link between immune cell in ltration and tumor expansion ( Figure 3A). We next characterized the accumulation of leukocyte subtypes at day 15. We found that CD11b + myeloid cells dominated the CD45 + in ltrate in both groups when compared to T, B lymphocytes and NK cells ( Figure   3B). Among CD11b + cells, we observed that CD11b + Ly6C + cells, which included a heterogeneous mix of CD11b + Ly6C hi Ly6Gand CD11b + Ly6C int Ly6G + cells signi cantly increased under HFHCD (Figure 3B,C).
A substantial increase in CD11b + Ly6G -F4/80 + macrophages was also observed in the HFHCD group in comparison to the CD group ( Figure 3B,C). Those results are recapitulated in a T-distributed stochastic neighbor embedding (t-SNE) representation ( Figure 3D).
As we showed that pro-atherogenic HFHCD signi cantly increased Ly6C hi monocyte numbers in blood ( Figure 1A), we tested the hypothesis that enrichment of CD11b + Ly6C + cells and macrophages in the tumor of mice under HFHCD was directly linked to the circulating monocytosis. We rst validated in our model that myeloid cells in the tumor derived, at least in part, from circulating blood cells. In vivo labeling of circulating monocytes and neutrophils was performed with intravenous injection of a combination of anti-CD115 plus anti-Ly6G in mice. After 6 hours, we found labeled monocytes in the CD11b + Ly6C hi population, and labeled neutrophils in the CD11b + Ly6C int population (Supplemental gure 2A). We next quanti ed monocyte in ltration in the tumor with uorescent latex microbeads as previously shown 24,30 .
Beads were injected intravenously in mice under CD and HFHCD at day 8, and tumors were harvested 24 hours later. Labeled-monocytes were found in the tumors of both groups and tended to accumulate more in tumors under HFHCD (supplemental gure 2B). Taken together, these results, which agree with the previous literature 31,32 , show that circulating monocytes and neutrophils supply a substantial part of the CD11b + Ly6C + pool in the forming tumor, and this process is increased in condition of HFHCD.
We next sought to investigate whether myeloid cell depletion would dampen tumor growth under HFHCD.
Repeated injections of clodronate liposomes were used to deplete phagocytes ( Figure 4A). Ly6C lo monocytes have been shown to exert higher phagocytosis capacity Clodronate liposomes depleted The loss of IL-1β or myeloid cell-derived VEGF-A recapitulates the bene cial effect of myeloid cell depletion on tumor development in mice fed a high fat/high cholesterol diet In addition to modulating immune cell production and in ltration, pro-atherogenic HFHCD has been shown to reprogram innate immune cell phenotype, both processes being thought to contribute to in ammatory disease progression 33 . As expected, we observed that LPS±IFNγ stimulated splenocytes from HFHCD treated-mice presented a robust pro-in ammatory response as compared to mice fed a normal CD ( Figure 5A). Then, we used a luminex assay to analyze the immune pro le of leukocytes in the tumor microenvironment of mice under CD or HFHCD. We found that CD45 + leukocytes isolated from the tumors of HFHCD-treated mice produced higher amounts of chemokines CCL5, CCL2, and CCL4 ( Figure   5B), which supported our ndings of increased myeloid cell recruitment in the tumor (see above). In addition, we found a signi cant increase of pro-in ammatory and pro-angiogenic markers TNF-α, IL-1β and VEGFR2 in conditioning media of leukocytes from the tumor of mice under HFHCD ( gure 5C). These data show the emergence of a systemic pro-in ammatory immune pro le in mice fed a HFHCD, which may trigger both systemic myelopoiesis and recruitment of pro-tumoral myeloid cells to the site of tumor development. We evaluated the direct role of tumor-associated immune cells on tumor cell proliferation using a co-culture assay. We isolated total CD45 + cells from tumors of mice on CD or HFHCD for 9 days and co-cultured them with fresh B16-F10 melanoma cells ex vivo. After 24 hours of co-culture, tritiated thymidine was added to the medium in order to quantify B16-F10 proliferation capacity. Co-culture with leukocytes isolated from tumors of mice under CD signi cantly altered the proliferation capacity of B16-F10 cells ( Figure 5D), showing that CD45+ cells can either limit tumor cell proliferation or kill tumor cells.
In contrast, when co-cultured with leukocytes isolated from tumors of mice under HFHCD, B16-F10 melanoma cells kept their high proliferation capacity ( Figure 5D), showing that CD45+ cells from a dyslipidemic environment favor tumor cell expansion.
We hypothesized that IL-1β might be an important actor of tumor growth acceleration in the context of a proatherogenic diet. We focused on IL-1β because we found it increased in tumor-associated leukocytes from HFHCD-fed mice ( gure 5C) and because it has long been established as a central regulator of in ammation, myelopoiesis 34 , vascular permeability and VEGF-mediated angiogenesis 35 , all processes being most likely solicited during tumorigenesis under HFHCD. The protective role of IL-1β de ciency in the development of experimental melanoma was previously shown 36 . IL-1 β-de cient mice and control mice were subjected to the HFHCD or CD as shown in gure 6A. Interestingly, we found that IL-1β de ciency completely protected mice from tumor growth acceleration under HFHCD ( Figure 6B). Because IL-1β has been shown to activate the generation and function of MDSCs, we rst tested the immunosuppressive capacity of IL-1β-de cient CD11b + Ly6C + cells isolated from tumors of mice under CD or HFHCD on sensitized OT-1 CD + 8 T cells, in vitro. We found that HFHCD increased the immunosuppressive capacities of CD11b + Ly6C + cells independently of IL-1β ( Figure 6C). This shows that CD11b + Ly6C + cells are analogous to myeloid-derived suppressor cells (MDSCs) and that their immunosuppressive functions are exacerbated under HFHCD. Because IL-1β appeared to be dispensable for the MDSC activity in our context, we tested whether it could control the generation and accumulation of MDSC in the tumors. Therefore, we monitored immune cell ratios in the blood and tumors of IL-1βde cient mice. We observed a strong decrease of Ly6C hi and Ly6C lo monocytes and neutrophils in the blood of IL-1β -/mice in comparison to IL-1β +/+ mice under HFHCD ( Figure 6D), suggesting that IL-1β is necessary for monocyte production and/or mobilization in this context. IL-1β de ciency also impaired MDSC in ltration into the tumors of mice under HFHCD ( Figure 6E). These results suggest that IL-1β production by immune cells controls HFHCD-associated monocytosis and subsequent accumulation of MDSCs in the tumor microenvironment, with no evidence for a direct effect on MSDC immunosuppressive activity against T cells.
Besides its direct function on myeloid cell maintenance, IL-1β has also been shown to control angiogenesis, in part by interacting with the VEGF/VEGFR pathway 37 . IL-1R signaling has been shown to induce VEGF expression and secretion in a number of cell types 38 , among which classical Ly6C hi monocytes 39 . Accordingly, we found a positive correlation between IL-1β release and VEGFR2 expression by tumor-isolated CD45 + cells ( Figure 7A). We therefore hypothesized that myeloid-derived VEGF-A may directly contribute to tumor growth expansion under HFHCD through direct effect on monocyte recruitment to the tumor and regulation of angiogenesis. As mentioned above, increased VEGF-A was found in tumors of mice under HFHCD. Besides tumor cells, stromal cells and particularly myeloid cells are thought to contribute to increased gradients of VEGF-A during tumorigenesis 40,41 . We used LysMCre +/-/VEGF f/f (VEGF-A ΔLysM ) mutant mice with a speci c deletion of VEGF-A in the myeloid cell compartment, and LysMCre -/-/VEGF f/f (WT) as control littermate mice. Mouse weight and cholesterol levels were unaffected by VEGF-A deletion in myeloid cells ( Supl Fig 4 A, B). We quanti ed the contribution of myeloid cells to the production of VEGF-A in the tumors. VEGF-A production was increased in tumors of WT mice fed a HFHCD, as expected ( Figure 7B). Interestingly, speci c deletion of VEGF-A in myeloid cells was su cient to prevent the increase of VEGF-A in tumors under HFHCD ( Figure   7B). This suggests that the increase in VEGF-A in tumors in response to HFHCD is mainly of myeloid cell origin. On the other hand, myeloid cells modestly contributed to the pool of VEGF-A in tumors of mice fed a normal CD ( Figure 7B). In line with this result, we found that the absence of myeloid-derived VEGF-A inhibited tumor growth ampli cation in response to HFHCD and recapitulated the effect of myeloid cell depletion, suggesting that the pro-tumoral activity of myeloid cells in the context of HFHCD is VEGF-Adependent ( Figure 7C). Under CD conditions, VEGF-A de ciency in myeloid cells tended to aggravate tumor progression in mice, in line with previous studies 20 , although it did not reach signi cance in our experimental setting. Importantly, we found a positive correlation between the levels of VEGF-A protein in the tumor and the weight of the tumor, con rming the importance of VEGF-A in tumor growth and emphasizing the crucial contribution of myeloid cells in this process ( Figure 7D). Mechanistically, we found that the absence of VEGF-A in myeloid cells had no impact on myeloid cell numbers in blood and in the tumor (Supl Fig 4C), which showed that the pro-tumoral effect of myeloid VEGF-A under HFHCD did not act directly on myeloid cell supply to the tumor. We found that VEGF-A de cient myeloid cells had an exacerbated in ammatory phenotype. Stimulated VEGF-A ΔLysM macrophages synthetized higher levels of IL-1β and lower levels of IL-10, in comparison to WT mice (Supl g 4D). This is in line with previous reports 42 and may explain the lack of protection against tumor development in CD conditions. The absence of VEGF-A in in amed macrophages did not in uence their inhibitory effect on B16-F10 proliferation, suggesting no direct effect of myeloid cells on tumor cell proliferation (Supl g 4E). However, in vivo VEGF-A disruption in myeloid cells stabilized tumor vessel perfusion, as assessed by echo Doppler imaging ( Figure 7E) and increased ICAM-1 total expression in the tumor in response to HFHCD ( Figure 7F). Taken together, these results show that in a dyslipidemic context, IL-1β increases the pool of myeloid cells with higher immunosuppressive and pro-angiogenic functions in tumors.

Discussion
In ammation is known to be increased in individuals with dyslipidemia and has become an interesting therapeutic target in oncology. The purpose of this work was to address the role of low-grade in ammation, as a result of dyslipidemia, on solid tumor growth. Here, we report that low-grade in ammation, caused by an atherogenic high fat/high cholesterol diet (HFHCD) and evidenced by increased monocyte level, accelerated the growth of solid melanoma in mice. Detailed phenotyping revealed that HFHCD-associated Ly6C hi monocytosis directly supplied myeloid-derived cell accumulation to the growing tumor, a process controlled by IL-1β. We found that in the HFHCD context, myeloid-derived suppressor cells favored tumor growth by sustaining the in ltration of myeloid cells with higher immunosuppressive capacities and increased VEGF-A production. In support of this interpretation, we found that myeloid cells rapidly dominated the tumor's in ltrate, and to a higher extent in the HFHCD-fed group compared to the CD-fed group. We demonstrated, with tracking techniques, that neutrophils and monocytes with MDSC characteristics were both recruited into the tumor. This is in line with studies showing that MDSCs arise from circulating monocytes and neutrophils, and does not rule out the possible accumulation of resident myeloid cells or circulating precursors 43,44 , which was not addressed in this study. We then used clodronate-liposomes depletion strategy to directly address the role of myeloid cells in tumor development in response to HFHCD. Repeated injection of clodronate liposomes did not deplete neutrophils but instead induced a neutrophilia both in mice fed with CD and HFHCD. This observation is in line with a recent publication 45 , and emphasizes the limitation of using clodronateliposomes treatment as experimental model for neutrophil depletion. Clodronate-induced neutrophilia did not increase tumor size, suggesting a minor role of neutrophils in melanoma growth in our model, at least upon monocyte depletion. Interestingly, under CD, clodronate liposomes decreased only circulating Ly6C lo monocytes, which were not associated with a reduction in tumor growth, showing that Ly6C lo monocytes did not contribute to tumor expansion. In contrast, under HFHCD conditions, clodronate liposomes additionally decreased Ly6C hi monocyte levels, which inhibited the growth of tumors. This demonstrates that Ly6C hi monocytes were direct actors of tumor development. Intriguingly, chronic depletion of Ly6C hi monocytes was more e cient under HFHCD than under CD conditions. This suggests an increased sensitivity of Ly6C hi monocytes to clodronate under HFHCD and an exhaustion of Ly6C hi monocyte precursors. Apart from a role of HFHCD on monocytosis, we found that HFHCD also increased the in ammatory pro le of leukocytes and the immunosuppressive capacities of MDSCs from the tumor microenvironment. Indeed, CD45 + cells isolated from the tumor of HFHCD-fed mice secreted higher levels of in ammatory cytokines, chemokines and growth factors, compared to cells from CD-fed mice. This suggests the development of a more in ammatory pro le of leukocytes educated in HFHCD conditions. This corroborates observations in obese patients suggesting that monocyte subsets increase their in ammatory phenotype and migration capacity 46 .
Beside its undisputed effect on myeloid cells, HFHCD could also affect tumor cells. We addressed the possibility of a direct effect of HFHCD on tumor cell metabolism by using FDG-PET-CT. We did not observe any increase in mean intensity of FDG consumption in the tumor, suggesting the absence of a direct effect of HFHCD on tumor cell glucose metabolism. However, it cannot be ruled out that HFHCD could have favored tumor cell proliferation through increased consumption of cholesterol and free fatty acids, or by epigenetic modi cations of oncogenes, as recently shown 47 .
The model of deleterious enrichment of pathogenic myeloid cells in in ammatory tissues through a surplus and increased supply of blood monocytes has been well documented in cardiovascular pathologies such as myocardial infarction, atherosclerosis and abdominal aortic aneurysm, where resolution of in ammation is altered 48-50 . Monocyte subsets are sequentially recruited to the in amed site, proportionally to their number in the circulation 51 . In tumorigenesis, even though the absolute count of monocytes in blood and the lymphocyte to monocyte ratio have strong prognostic value in various forms of cancer 52 , the contribution of excessive monocyte production to tumor growth during dyslipidemia is unknown. Nevertheless, a very recent publication demonstrated that myocardial infarction, a clinical manifestation of dyslipidemia, accelerates breast cancer in mice through increased circulating Ly6C hi monocyte levels and recruitment to tumors and through reprogramming of these monocytes to an immunosuppressive phenotype 53 . Our study corroborates these ndings and reinforces the in ammatory link between cardiovascular pathologies and tumorigenesis. A few studies previously pointed out the effect of Western diet on tumorigenesis in C57BL/6J mice, but they focused on tumor cells, not on host immune cells 54,55 . In one study, the use of a diet similar to that used in our study accelerated experimental prostate tumor growth, which was attributed to increased cholesterol uptake by the proliferative tumor 54 . That study emphasized the potential bene t of managing cholesterol metabolism as an anti-tumoral approach. Several studies have been conducted, since then, to further assess the association between statin use and mortality in different types of cancer, but there is no consensus regarding the molecular targets of the potential anti-cancer effects of statins 56 . On the other hand, the recent emergence and e cacy of immunotherapy, in particular in melanoma, as an arti cial stimulation of the immune system to ght cancer cells, has led to the need of understanding how immune cells are educated in the systemic and local pro-tumoral environments. Even more recently, a growing appreciation of the potential contributions of innate immune effectors to anti-tumor immunity, especially in the context of combination immunotherapy, is leading to novel strategies to elicit a more integrated immune response against cancer. Since innate cells are central actors of the in ammatory response, it is most likely that the contribution of innate cells on tumorigenesis will be further emphasized during chronic in ammatory conditions. In our present work, we propose for the rst time the concept that silent mild-dyslipidemia without obesity creates an environment that is conducive for the accumulation of pro-tumoral myeloid cells, which facilitates tumor growth. This study paves the way to new opportunities in the eld of immunotherapy, targeting Ly6C hi monocytes in individuals with known low-grade in ammation, in particular in the context of mild-dyslipidemia. The importance of monocyte subsets was recently pointed out in several publications, demonstrating their importance in cancer therapy. For instance, classical Ly6C hi monocytes were proposed to predict response to anti-PD-1 immunotherapy 57 , and non-classical Ly6C lo monocytes were thought to contribute to resistance against anti-VEGF therapies in mouse models of colorectal cancer 58 . These discoveries might even be more relevant in the context of low-grade chronic in ammation, as we found a predominant role of monocytes on tumor growth in this particular context. Importantly, our data showed that VEGF-A expression was boosted in tumors of mice on HFHCD, and that was due to excessive myeloid accumulation in the tumor microenvironment at early time points. Our results bring new elements regarding the underexplored role of myeloid-derived VEGF-A in in ammatory diseases. Myeloid-derived VEGF-A was shown to regulate angiogenesis in mouse models of lung injury 59 and melanoma 20 . Alternatively, myeloid-derived VEGF-A was shown to control in ammation independently of vasculogenic effects in granulomatous in ammation 60 and in myocardial infarction 42 . A key question raised by our study is whether IL-1β and VEGF-A contributed to tumor progression in a sequential and stage-speci c manner, or in a synergic way. In the context of low-grade in ammation, our results show that inhibition of either IL-1β or VEGF-A recapitulates the anti-tumoral effect of classical monocyte depletion. This suggests the sequential contribution of IL-1β and VEGF-A in early tumorigenesis in this context. In addition, our ndings suggest a negative control of VEGF-A on IL-1β production in myeloid cells, which would indicate the absence of synergistic effects. Importantly, tumor cells themselves appear as little contributors of IL-1β and VEGF-A production in early metabolic syndrome, which shows that systemic dysfunction and plasticity of the immune environment dictate the early steps of tumorigenesis. While IL-1β would maintain the production of myeloid cells and probably stimulate their production of VEGF-A, VEGF-A would shape the tumor microenvironment. Complementary upregulated cytokines and chemokines may participate in this process. In addition to modulating blood ow, we found that VEGF-A controlled the phenotype of myeloid cells. As we found that MDSCs from HFHCD fed mice presented a higher capacity of immunosuppression, it would be interesting to evaluate the link, in conditions of chronic low-grade in ammation, between VEGF-A and the expression of inhibitory checkpoints on immune cells, as VEGF-A produced in the tumor microenvironment was recently shown to enhance expression of PD-1 and other inhibitory checkpoints involved in CD8 + T cell exhaustion 61 . In view of these results, association of anti-angiogenic molecules or statins with immunomodulators of inhibitory checkpoints may be of particular interest in solid tumors in individuals with pre-metabolic syndrome status.

Declarations Fundings
This work was supported by grants from the ARC ("Association pour la recherche sur le cancer") and the "Ligue contre le cancer" foundations. Tartour's laboratory is supported by the "Ligue contre le cancer" foundation.  Total numbers of immune cell in ltration were showed (n=4-7 mice/group from one representative experiment). C) Gating strategy for myeloid cells in tumors D) Single live CD45 cells from the ow cytometry data of tumors were concatenated and subjected to the t-SNE dimensional reduction. CD11b expression in CD or HFHCD tumors are colored mapped from blue (low expression) to red (high expression) into the t-SNE map. CD11b+Ly6Chi, CD11b+Ly6CintLy6G+, CD11b+F4/80+ populations were overlayed into the t-SNE map. Mean±sem. Mann-Whitney t-test. *p<0.05  HFHCD educates leukocytes to become pro-in ammatory in systemic and in the tumor microenvironment 0,25x106 B16-F10 cells were implanted subcutaneously into C57BL6/J mice, fed with CD or HFHCD for 2 weeks. Diet was maintained until the end of experiment. A) Splenocytes from CD and HFHCD-fed mice were stimulated in vitro with LPS (10υg/ml) +/-IFNγ (100UI/ml) for 24 hours. Cytokine secretion was measured in the supernatant by ELISA. Unpaired t-test. *p<0.05, * * p<0.01 B-C) CD45+cells were isolated from B16-F10 tumors at day 15 post-injection. And cultured in medium for 24h. Chemokines (B), in ammatory and angiogenic factors (C) were measured in the medium by ELISA or Luminex assay (n=4/group). D) B16-F10 cell line (10 000 cells/well) was co-cultured with CD45 cells (2 500 cells/wells) for 24h. B16-F10 proliferation was measured by 3H-Thymidine incorporation (n=3-5/group). Mean±sem.