Hypoxia Drives Dihydropyrimidine Dehydrogenase Expression in Macrophages

bioRxiv preprint doi: https://doi.org/10.1101/2020.10.15.341123; this version posted October 15, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

1 Hypoxia Drives Dihydropyrimidine Dehydrogenase Expression in Macrophages

2 and Confers Chemoresistance in Colorectal Cancer

4 Marie Malier1,2, Magali Court1,2, Khaldoun Gharzeddine1,2,3, Marie-Hélène Laverierre1,2,4, Sabrina 5 Marsili5,6, Fabienne Thomas5,6, Thomas Decaens 2,7,8, Gael Roth 2,7,8, Arnaud Millet1,2,3

6 1. Team Mechanobiology, Immunity and Cancer, Institute for Advanced Biosciences Inserm 1209 – 7 UMR CNRS 5309, Grenoble, France

8 2. Grenoble Alpes University, Grenoble, France

9 3. Research department, University Hospital Grenoble-Alpes, Grenoble, France

10 4. Department of pathological anatomy and cytology, University Hospital Grenoble-Alpes, Grenoble, 11 France

12 5. CRCT Inserm U037, Toulouse University 3, Toulouse, France

13 6. Claudius Regaud Institute, IUCT-Oncopole, Toulouse, France

14 7. Department of Hepatogastroenterology, University Hospital Grenoble-Alpes, Grenoble, France

15 8. Team Tumor Molecular Pathology and Biomarkers, Institute for Advanced Biosciences UMR 16 Inserm 1209 – CNRS 5309, Grenoble, France

18 Corresponding author:

19 Arnaud Millet MD PhD

20 Team Mechanobiology, Immunity and Cancer, Institute for Advanced Biosciences

21 Batiment Jean Roget 3rd floor

22 Domaine de la Merci

23 38700 La Tronche

24 France

25 e-mail: [email protected]

27 Abstract

28 Colon adenocarcinoma is characterized by an infiltration of tumor-associated macrophages (TAMs). 29 TAMs are associated with a chemoresistance to 5-Fluorouracil (5-FU), but the mechanisms involved 30 are still poorly understood. In the present study, we found that macrophages specifically overexpress 31 dihydropyrimidine dehydrogenase (DPD) in hypoxia, leading to a macrophage-induced 32 chemoresistance to 5-FU by inactivation of the drug. Macrophage DPD expression in hypoxia is 33 translationally controlled by the cap-dependent protein synthesis complex eIF4FHypoxic, which includes 34 HIF-2α. We discovered that TAMs constitute the main contributors to DPD activity in human colorectal 35 primary or secondary tumors where cancer cells do not express significant levels of DPD. Together, 36 these findings shed light on the role of TAMs in forming chemoresistance in colorectal cancers and 37 offer the identification of new therapeutic targets. Additionally, we report that, contrary to what is 38 found in humans, macrophages in mice do not express DPD.

40 Keywords: Hypoxia, dihydropyrimidine dehydrogenase, tumor-associated macrophages, 41 chemoresistance, colorectal cancer, HIF-2α.

58 Introduction

59 Colorectal cancers (CRC) are a leading cause of death worldwide and constitute the third cancer- 60 related cause of death in the United States (Siegel et al., 2020). Chemotherapy is one of the tools used 61 to treat these tumors, however some patients do not respond well to this treatment, resulting in poor 62 prognosis. This chemoresistance is caused by various mechanisms such as drug inactivation, drug 63 efflux from targeted cells and modifications of target cells (Marin et al., 2012; Vasan et al., 2019). 64 Interestingly, the importance of tumor microenvironment in chemoresistance has recently garnered 65 attention. The tumor immune microenvironment (TIME), notably through its innate immune part that 66 is mainly composed of tumor-associated macrophages (TAMs), deserves particular attention 67 (Binnewies et al., 2018; Ruffell and Coussens, 2015). TAMs have been associated with bad prognosis 68 in the case of various solid tumors (Yang et al, 2018) and have been shown to orchestrate a defective 69 immune response to tumors (DeNardo and Ruffell, 2019). It has been suggested that TAMs are 70 reprogrammed by cancerous cells to secondarily become supporting elements of tumor growth (Aras 71 and Zaidi, 2017). The involvement of TAMs in CRC has been controversial, and only recently has their 72 association with a poorer prognosis been recognized (Pinto et al., 2019; Ye et al., 2019). Their 73 implication in chemoresistance, particularly against 5-Fluorouracil (5-FU) a first line chemotherapy in 74 CRC has been reported (Yin et al., 2017; Zhang et al., 2016), suggesting that macrophage targeting 75 could facilitate a way for increasing treatment efficiency. However, the precise mechanisms by which 76 TAMs participate in creating chemoresistance in human colorectal tumors are still unclear. 77 Interestingly, TAMs increase hypoxia in tumor tissues, which is susceptible to favoring 78 chemoresistance in return (Jeong et al., 2019). We also know that hypoxia affects macrophage biology 79 (Court et al., 2017) and could mediate resistance to anticancer treatment and cancer relapse (Henze 80 and Mazzone, 2016). Based on the abundance of macrophages in CRC, we hypothesized that hypoxia 81 could directly modulate macrophage involvement in 5-FU resistance.

82 Results

83 Macrophages confer a chemoresistance to 5-FU in a low-oxygen environment

84 In order to evaluate the effect macrophages in the vicinity of the tumor have on chemotherapy, we 85 analyzed the impact on cancer cells growth of conditioned medium (CM) by macrophages containing 86 5-fluorouracil (5-FU) or none, designed as CM(5-FU) and CM(Ø) respectively. To examine the role of

87 oxygen, we performed these experiments in normoxia (N=18.6% O2) and in hypoxia (H=25mmHg ~3%

88 O2). Non-conditioned 5-FU strongly inhibited HT-29 and RKO proliferation independently of oxygen 89 concentration (Figures 1A and 1B). CM(Ø) had little effect on the proliferation of colon cancer cells 90 (Figures 1A and 1B). Meanwhile, CM(5-FU 1 µg/mL) inhibited cell growth in normoxia but macrophage

91 conditioning in hypoxia provided complete protection against 5-FU inhibition (Figure 1A, 1B). RKO cells 92 were found sensitive to a lower concentration of 5-FU (0.1 µg/mL). In this state, additionally 93 macrophage conditioning was able to protect against 5-FU not only in hypoxia but also in normoxia 94 (Figure 1B). This observation led us to consider an inactivation mechanism of 5-FU driven by 95 macrophages with increased efficiency in hypoxia. A previously proposed mechanism for macrophage- 96 induced chemoresistance in CRC was related to their ability to secrete interleukin (IL)-6, leading to an 97 inhibition of cancer cell apoptosis (Yin et al., 2017). We first try to verify wether we can confirm the 98 presence of IL-6 in our CM by human monocyte-derived macrophages in normoxia and in hypoxia and 99 found no spontaneous secretion (<10 pg/mL) of IL-6 (data not shown). As the conditioning by 100 macrophages provided a complete protection, we then hypothesized that a direct action of 101 macrophages on 5-FU was the likely mechanism. In order to obtain a molecular explanation of the 102 differing effect under various oxygen concentrations, we performed a proteomic analysis of human 103 macrophages in hypoxia compared to normoxia. Our quantitative proteomic approach revealed that 104 dihydropyrimidine dehydrogenase (DPD) is strongly overexpressed in hypoxia (Figure 1C). DPD (coded 105 by the DPYD gene) is the rate-limiting enzyme of the pyrimidine degradation pathway. DPD adds two 106 hydrogen atoms to uracil, with NADPH as an obligatory cofactor, leading to dihydrouracil, which is 107 secondarily degraded to β-alanine under the control of DPYS and UPB1 (Figure 1D). 5-FU is a 108 fluorinated, analogue of uracil reduced by DPD to 5-fluorodihydrouracil in an inactive compound 109 (Figure 1E). We confirmed through immunoblotting the increased expression of DPD in hypoxia when 110 compared to the basal expression in normoxia (Figure 1F). Since the putative action of DPD on 5-FU is 111 related to its enzymatic activity, we also confirmed that DPD expression in hypoxic macrophages was 112 functional, leading to an increase of the dihydrouracil to uracil ratio in the extracellular medium 113 (Figure 1G).

114 Chemoresistance to 5-FU is driven by increased DPD activity in hypoxic macrophages

115 In order to confirm the inactivation of 5-FU by macrophage DPD and its potential clinical relevance, 116 we analyzed the kinetics of 5-FU degradation in normoxia and hypoxia. As 5-FU is a small molecule, its 117 diffusion in tissue is quite high. Indeed, following four days of oral ingestion of 5-FU the plasmatic 118 concentration was found ~ 1 µg/mL (Zheng and Wang, 2005). The mean 5-FU concentration was 0.411 119 ± 0.381 (μg/g of the tissue) in the tumor portions of the specimens and 0.180 ± 0.206 (μg/g of the 120 tissue) in the normal portions in colorectal cancers (Tanaka-Nozaki et al., 2001). As TAMs represent 2 121 -10% of cells in CRC, especially localized in the invasion front (Pinto et al., 2019) and since that 1g of 122 tissue typically contains ~ 108 cells (Wilson et al., 2003), a reasonable estimated ratio in CRC is ~ 106 123 macrophages/g of tissue. According to 5-FU reported concentration, an estimated physiologically 124 relevant ratio in CRC is 1µg of 5-FU/106 macrophages. We interestingly found that 2 µg of 5-FU could

125 be eliminated by 106 macrophages in hypoxia in less than 24 h and that normoxic macrophages were 126 unable to completely eliminate this quantity in 48 h (Figure 2A). We validated that 5-FU degradation 127 was due to DPD catalytic activity using gimeracil, which is a specific inhibitor of DPD (Figures 2B and 128 2C). Non-conditioned 5-FU induced death in HT-29 or RKO irrespective of the presence of oxygen and 129 gimeracil, demonstrating no significant DPD activity in these cells (Figures 2D and 2F). We found no 130 significant level of expression of DPD at the protein level, neither in HT-29 nor in RKO cells, irrespective 131 of the oxygen concentration (Figures S1A and S1B). Furthermore, we found a downregulation of the 132 DPYD mRNA in HT-29 and in RKO, emphasizing a defective transcription of the DPD gene (Figure S1C). 133 We then confirmed that the pharmacological inhibition of the methyltransferase EZH2 by the specific 134 inhibitor GSK126 led to a detectable level of DPYD mRNA in RKO cells (Figure S1D). It has been reported 135 that the transcription factor PU.1 drives the expression of DPYD and that EZH2 is responsible for the 136 histone H3K27 trimethylation at the DPYD promotor site leading to its downregulation in colon cancer 137 cells (Wu et al., 2016). Whereas CM(5-FU 1µg/mL) induced cell death in HT-29 cells in the case of 138 normoxia, its cytotoxic effect dramatically decreased in hypoxia and this could be reverted by 139 inhibition of DPD activity using gimeracil (Figure 2D). Chemoresistance appeared to be based solely 140 on DPD activity promoted by hypoxia. We observed a similar result using a 3D tumor model growth 141 of HT-29 exposed to CM (Figure 2E), and we confirmed the generality of this mechanism in RKO cells 142 (Figure 2F). In order to confirm that oxygen was the main factor controlling DPD expression in hypoxia, 143 we verified that DPD was not induced or repressed by 5-FU itself (Figure 2G). We also explored wether 144 cancerous cells can modulate DPD expression in macrophages. Using a transwell co-culture between 145 cancerous cells and macrophages revealed no modulation in DPD expression in macrophages (Figure 146 2H). In order to assess the efficiency of DPD degradation of 5-FU, we also performed a direct co-culture 147 assay between cancerous cells and macrophages and found that macrophages protected cancerous 148 cells from 5-FU in a DPD dependent manner (Figure 2I).

149 Oxygen controls DPD expression in macrophages translationally through the cap-dependant protein 150 synthesis molecular complex eIF4FHypoxic

151 We discovered that a decreased oxygen concentration was able to increase the expression of DPD in 152 human macrophages. In order to gain further insight into this, we carried out the transition of 153 macrophages to various oxygen environments, to study the way DPD is controlled. We observed that 154 DPD expression was inversely correlated to oxygen levels during the transitions (Figures 3A and 3B). 155 The evolution of DPD expression was then analyzed with the help of a first-order differential equation 156 (Figure 3C). This analysis revealed that the DPD half-life increased during the HN (Hypoxia to 157 Normoxia) transition, demonstrating no increased degradation of DPD (Figure 3C). These results 158 sugggested a synthesis control of DPD expression rather than a degradation control. We then

159 demonstrated that hypoxic transitions were associated with the production of a functional DPD 160 resulting in an increased dihydrouracil/uracil ratio in the extracellular medium (Figure 3D). Besides, 161 we noted that profound hypoxic conditions (7 mmHg ~ 1% O2) provided the same induction of DPD 162 overexpression as moderate hypoxic conditions (Figure 3E). As Hypoxic-induced factor HIF-1α is 163 known to be stabilized during hypoxic transitions, so we checked whether its stabilization could be 164 implicated in DPD over-expression. We found no induction of DPD synthesis due to HIF-1α stabilization 165 in human macrophages (Figure 3F). We next sought to determine whether the expression of DPD is 166 transcriptionally controlled when macrophages are exposed to low oxygen environments. To do so, 167 we analyzed mRNA levels of oxygen-sensitive genes in macrophages (VEGF-A, NDRG1, P4HA1, SLC2A1) 168 in the transition from normoxia to hypoxia or from hypoxia to normoxia. We discovered that the DPYD 169 mRNA level did not present any significant variation of its level of expression compared to oxygen 170 responsive genes (Figure 3G). We confirmed the absence of transcriptional control by inhibiting the 171 synthesis of new mRNAs with actinomycin D and found no effect on DPD protein synthesis during a 172 hypoxic transition (Figure 3H). This absence of correlation between the mRNA level and protein 173 expression suggested a translational controlled mechanism. Recently, it has been demonstrated that 174 the initial steps of protein synthesis such that as binding of the eukaryotic translational initiation factor 175 E (eIF4E), part of the eIF4FNormoxic initiation complex, to mRNA are repressed in hypoxia. Another 176 complex, eIF4FHypoxic, involving eIF4E2, RBM4 (RNA binding protein 4), and HIF-2α interacts with mRNA 177 and mediates a selective cap-dependent protein synthesis in low oxygen environments (Ho et al., 178 2016; Uniacke et al., 2012). Genes containing a RNA hypoxic response element (rHRE) in their 3’UTR 179 could bind to the HIF2α-RBM4-eIF4E2 complex part of eIF4FHypoxic (Uniacke et al., 2012). Therefore, we 180 verified if DPYD presents such rHRE sequence on its 3’UTR and found effectively a trinucleotide CGG 181 RBM4 binding motif (Figure 3I). Using these results, we depleted the expression of eIF4E and eIF4E2 182 in macrophages using specific siRNAs and found that hypoxia induced DPD synthesis is eIF4E2 183 dependent (Figure 3J). We also confirmed the involvement of HIF-2α in DPD synthesis in hypoxia 184 (Figure 3K). These results demonstrated that DPD expression is controlled by the translation complex 185 eIF4FHypoxic when oxygen decreases without involving a transcriptional regulation. Interestingly, the 186 participation of HIF-2α in the eIF4FHypoxic complex is independent of its transcription factor activity 187 (Uniacke et al., 2012). We have demonstrated that HIF-2α stabilization during hypoxia is the driving 188 sensing signal that leads to the production of DPD in macrophages.

189 TAMs in human colon cancer tissues harbor the principal component of DPD expression in tumors

190 We have demonstrated that DPD expression in macrophages confers a chemoresistance to 5-FU. In 191 der to assess the clinical relevance of this result, we further determined the relative expression of DPD 192 in various cellular populations found in colorectal tumors and colorectal liver metastasis. Based on

193 immunohistochemical analysis of CRC tissues, it had been previously reported that cancer cells do not 194 express DPD whereas normal cells, morphologically similar to macrophages, present a strong 195 expression (Kamoshida et al., 2003). In order to appreciate the quantitative effect of this expression 196 in macrophages in CRC, we analyzed the corresponding level of DPD expression in colorectal cancerous 197 cells. We first used RNAseq analysis in various cancer cell lines from the Cancer Cell Line Encyclopedia 198 (CCLE) and confirmed that the 59 cancer cell lines originating from colon cancer presented the lowest 199 level of expression for DPYD suggesting a low expression pattern in these tumors (Figure 4A). This 200 result confirmed what we had observed for three colon cancer cell lines RKO, HT-29 (Figure S1A), and 201 Lovo (data not shown) and emphasized a preeminent role of macrophages in DPD induced 202 chemoresistance in tumors. We further analyzed tissue samples from patients suffering from 203 colorectal cancer with liver metastasis and found that the strongest expression of DPD was found in 204 areas with CD68+ macrophages (Figure 4B). Tumor cells did not present a significant level of DPD 205 expression in metastasis when compared to neighbouring TAMs (Figure 4B). Furthermore, in primary 206 tumors, macrophages were also found to express the highest level of DPD, but no detectable 207 expression was found in cancerous cells (Figure 4C). We further confirmed the exclusive expression of 208 DPD by macrophages, by showing that strongly DPD+ cells were also CD68+ using an 209 immunofluorescence co-expression analysis both in liver metastasis where (Figure 4E) and primary 210 tumors (Figure 4F). Because CD68 has been found to be less specific than previously thought as a 211 macrophage marker (Ruffell et al., 2012), we confirmed our results using the CD163 macrophage 212 marker. We confirmed that DPD+ cells are CD163+ TAMs in liver metastasis and primary tumors 213 (Figures S2A and S2B). Since macrophages can present various states of activation depending on their 214 surrounding environment (Sica and Mantovani, 2012), we tested if DPD could be influenced by 215 cytokines or growth factors. In this respect, we found that DPD expression in macrophages is not 216 modulated according to the polarization state induced by IL-4, IL-6, IL-10, IL-13, IFNγ or GM-CSF, in 217 contrast with the expression induced by hypoxia during the same period of time (Figure S2C). All these 218 results indicate that DPD expression in colorectal cancers at the primary site and liver metastasis 219 belong to macrophages, under the control of oxygen.

220 Rodents’ macrophages do not express significant levels of DPD

221 In order to assess the generality of the oxygen control of DPD expression in macrophages we checked 222 if this mechanism still holds in rodents. Surprisingly, we found that mice bone marrow derived 223 macrophages (BMDM) do not express a significant level of DPD in normoxia or hypoxia despite the 224 presence of the protein in mice liver (Figure 5A). No detectable level of DPYD mRNA was found in 225 BMDM from BALB/c mice (data not shown). We discovered a similar result in the RAW264.7 mice 226 macrophage cell line, where no protein (Figure 5B) or mRNA of DPYD was found (data not shown).

227 Similarly, we also found that chemically mobilized peritoneal macrophages from adult female Fischer 228 rats do not express DPD (data not shown). We then used open data sets from microarrays to compare 229 the expression of DPYD mRNA levels in humans to those in mice and found that DPYD presents the 230 highest levels of expression in human monocyte/macrophages from various anatomical 231 compartments, contrary to what was found in mice where macrophages expressed few mRNA DPYD 232 molecules compared to other cellular compartments (Figure 5C). This result suggests repression of 233 mRNA synthesis in mice macrophages. We confirmed the epigenetic control of gene expression using 234 5-aza-2’deoxycytidine (decitabine) a DNA hypo-methylation agent. Indeed, decitabine led to a strong 235 increase in mDPYD mRNA level of expression in treated RAW macrophages compared to non-treated 236 cells (Figure 5D). This result suggests that the methylation of the DPYD promoter gene in mice is, at 237 least in part, responsible for its repressed expression.

238 Transduced human DPYD in mice macrophages leads to 5-FU chemoresistance in vivo

239 In order to obtain a mice model mimicking the human DPD expression in macrophages, we transduced 240 the human DPYD gene incorporated into a lentivirus to obtain “DPD-humanized” mice macrophages 241 (Figures 5E and 5F). CM of mice macrophages expressing DPD were able to confer chemoresistance to 242 CT-26 (a mice colon cancer cell line that does not express DPD) demonstrating the functionality of the 243 transduced DPYD (Figure 5G). We also discovered that wild-type macrophages are associated with 244 weak protection toward 5-FU compared to macrophages expressing DPD, demonstrating that even if 245 other chemoresistance mechanisms are associated with TAMs DPD-induced chemoresistance is 246 probably the more efficient one (Figure 5G). In order to further confirm the relationship between DPD 247 expression in macrophages and chemoresistance in colorectal cancers, a tumor assay in mice was 248 performed. CT-26 and RAW macrophages expressing or not human DPD were implanted into flanks of 249 BALB/c mice. Ten days after the implantation, 5-FU at 25mg/kg was injected intraperitoneally during 250 five days for two consecutive weeks (Figure 5H). We confirmed that tumors harboring macrophages 251 expressing DPD were more resistant to 5-FU than the control tumors with wild-type macrophages 252 (Figures 5I and 5J), indicating that DPD expression in TAMs promotes chemoresistance in vivo.

253 Discussion

254 In recent years, the function of the immune system has become a key element in understanding tumor 255 interaction with the surrounding healthy tissues as well as a provider of new therapeutic strategies. 256 The tumor immune microenvironment (TIME) is composed of various types of immune cells; 257 nevertheless, TAMs usually represent quantitatively the largest population found in solid cancers. 258 TAMs are involved in tumor growth, immune evasion, neoangiogenesis and treatment resistance. 259 Using depletion methods, a large number of studies have reported an increased chemo-sensitivity

260 when macrophages are removed from the tumor (Ruffell and Coussens, 2015). Futhermore, co-culture 261 studies have revealed a macrophage-mediated resistance to various anti-cancer drugs such as 262 paclitaxel, doxorubicin, etoposide or gemcitabine (Mitchem et al., 2013; Shree et al., 2011). 263 Specifically, depletion of MHCIIlo TAMs leads to an increased sensitivity to taxol-induced DNA damage 264 and apoptosis (Olson et al., 2017). Another key point is that TAMs were mainly found in hypoxic areas 265 where they could further favor hypoxia by secreting VEGF-A, leading to the formation of an abnormal 266 vasculature (Murdoch et al., 2008). Accordingly, mechanisms involving macrophages-induced 267 chemoresistance relied on an indirect effect due to the secretions by macrophages such as pyrimidine 268 nucleosides (deoxycytidine) inhibiting gemcitabine induction of apoptosis in pancreatic ductal 269 adenocarcinoma (Halbrook et al., 2019). More specifically in CRC, the implication of macrophages in 270 chemoresistance has also been suggested based on in vitro and in vivo studies. The mechanisms 271 proposed were diverse but involved indirect secreted factors. For example, it has been proposed that 272 IL-6 secreted by macrophages could stimulate STAT3 in cancerous cells inducing the inhibition of the 273 RAB22A/BCL2 pathway through miR-204-5p expression, thereby leading to chemoresistance to 5-FU 274 (Yin et al., 2017). Similarly the secretion of putrescin by macrophages, a member of the polyamine 275 family, was shown to suppress the JNK/Caspase 3 pathway in cancerous cells, providing a protection 276 against 5-FU (Zhang et al., 2016).

277 To order to understand the involvement of TAM in chemoresistance in CRC, we designed the present 278 study to incorporate oxygen concentration as a key environmental parameter. We had previously 279 reported that the oxygen disponibility in macrophages’ environment greatly influences their immune 280 functions such as their ability to clear apoptotic cells (Court et al., 2017). As colon tissues are naturally

281 exposed to levels of oxygen that are usually lower than 5% O2 (Keeley and Mann, 2018) with values

282 that could reach even lower (<1% O2) in tumors, oxygen appeared as a fundamental parameter to 283 understand macrophage involvement in chemoresistance. We first verified if a secreted factor by 284 human macrophages could provide a chemoresistance and finally found a direct effect of hypoxic 285 macrophages on 5-FU (Figures 1A and 1B). A molecular analysis revealed that the dihydropyrimidine 286 deshydrogenase (DPD), an enzyme of the pyrimidine catabolism pathway, was over-expressed in 287 hypoxic human macrophages providing a direct chemoresistance mechanism (Figures 1C and 2). We 288 found that DPD was part of the protein synthesis response to hypoxia in macrophages under the 289 control of the translation initiating complex eIF4FHypoxic (Figure 6), notably composed of HIF-2α that is 290 stabilized in hypoxia but acts in this complex independently of its transcription factor activity (Uniacke 291 et al., 2012). Accordingly, we found that this mechanism was independent of polarization states, as 292 various polarizing factors were unable to modify DPD expression contrary to hypoxia (Figure S2B). This 293 result demonstrated the robustness of the oxygen control of DPD expression, irrespective of the

294 polarization states. As TAM populations have been demonstrated to be highly heterogeneous 295 according to their phenotype, which does not fit the classical M1/M2 classification (Aras and Zaidi, 296 2017), the universality of the mechanism identified in human macrophages assured its relevance in 297 colorectal cancers (Figures 4B to 4F and S2A).

298 DPD expression in macrophages seemed to be particularly relevant in CRC where cancerous cells 299 present a low expression level of the protein in primary tumors as well as in liver metastasis (Figures 300 4A, 4C and 4D), assuring macrophages to be mainly responsible for the tumor tissue degradation of 301 5-FU. This general feature seems to rely on the epigenetic control of DPD expression in cancer cells. 302 Indeed, many colon cancer cells lines have been noted to harbor a histone H3K27me3 mark that blocks 303 the fixation of the transcription factor PU.1, leading to the inhibition of DPD mRNA transcription (Wu 304 et al., 2016).

305 We further found that macrophages in rodents do not express DPD as it is transcriptionally repressed. 306 This finding forced us to re-evaluate previous in vivo models that are used to assess the involvement 307 of macrophages in CRC. Indeed, we have shown the importance of DPD activity in macrophages and 308 found that it represents the main quantitative source of degradation of 5-FU in human colorectal 309 tumors. In order to demonstrate the relevance of this mechanism to chemoresistance, we designed 310 an in vivo model using mice macrophages expressing DPD. We validated the importance of DPD 311 expression in TAMs leading to chemoresistance to 5-FU. Supporting these results, previous clinical 312 studies have suggested that DPD mRNA expression is a marker of chemoresistance in CRC. These 313 studies usually assumed that DPD expression is mainly caused by cancerous cells. Nevertheless, the 314 mRNA level in the tumor is correlated with a low response to 5-FU confirming its relevance (Ichikawa 315 et al., 2003; Salonga et al., 2000; Shirota et al., 2002; Soong et al., 2008). The results we have obtained 316 in our study suggest that the main prognostic factor for 5-FU response is DPD expression in 317 macrophages located in tumors and liver metastasis. That expression is notably important in the 318 invasive front, where TAMs seem to concentrate (Pinto et al., 2019). Furthermore, the invasion front 319 is known to be a hypoxic area in CRCs (Righi et al., 2015). Since the mechanism identified in this study 320 relied on quantitative expression of DPD by macrophages, the assessment of the spatial heterogeneity 321 of DPD expression will be necessary to stratify patients in various response groups for chemotherapy 322 (Marusyk et al., 2020). Thus this study constitutes an important advance in the understanding of the 323 role of tumor immune environment in chemoresistance to 5-FU in CRC. Finally, further translational 324 and clinical studies are needed to confirm the clinical relevance of these findings and develop new 325 predictive markers of response to 5-FU-based treatments in order to improve colorectal patients’ care.

326

327 MATERIALS AND METHODS

328 Cell culture

329 RAW264.7, CT-26, RKO, HT-29, Lovo were purchased from ATCC. RAW were maintained in high- 330 glucose DMEM (Gibco) supplemented with 10% FBS (Gibco) at 37°C, CT-26 and RKO were maintained 331 in RPMI (Gibco) supplemented with 10% FBS (Gibco) at 37°C, HT-29 were maintained in McCoy’s 332 medium (Gibco) supplemented with 10% FBS (Gibco) at 37°C and Lovo were maintained in F12-K 333 supplemented with 10% FBS (Gibco) at 37°C. All cells were routinely tested for mycoplasma 334 contamination using MycoAlert detection kit (Lonza). All cells have been used in the following year of 335 their reception.

336 Human samples

337 Human blood samples from healthy de-identified donors are obtained from EFS (French national blood 338 service) as part of an authorized protocol (CODECOH DC-2018–3114). Donors gave signed consent for 339 use of their blood in this study. Tumor samples were obtained from the department of pathology of 340 the university hospital of Grenoble as part of a declared sample collection AC-2014-2949. Patient 341 selection criteria were a diagnostic of colorectal adenocarcinoma and tissue samples availability for 342 the primary tumor and liver metastasis. Clinical characteristics of the patients are reported in the 343 table. All patients gave their signed consent for this study as part of an authorized protocol (INDS 344 MR0709280220).

345 Animals

346 8 weeks old Balb/c female mice were obtained from Charles River. Animals were housed and bred at 347 « Plateforme de Haute Technologie Animale (PHTA) » UGA core facility (Grenoble, France), EU0197, 348 Agreement C38-51610006, under specific pathogen–free conditions, temperature-controlled 349 environment with a 12-h light/dark cycle and ad libitum access to water and diet. Animal housing and 350 procedures were conducted in accordance with the recommendations from the Direction des Services 351 Vétérinaires, Ministry of Agriculture of France, according to European Communities Council Directive 352 2010/63/EU and according to recommendations for health monitoring from the Federation of 353 European Laboratory Animal Science Associations. Protocols involving animals were reviewed by the 354 local ethic committee “Comité d’Ethique pour l’Expérimentation Animale no.#12, Cometh-Grenoble” 355 and approved by the Ministry of Research under the authorization number (January 2020) 356 APAFIS#22660-2019103110209599.

357 Human macrophage differentiation from monocytes

358 Monocytes were isolated from leukoreduction system chambers of healthy EFS donors using 359 differential centrifugation (Histopaque 1077, Sigma) to obtain PBMCs. CD14+ microbeads (Miltenyi 360 Biotec) were used to select monocytes according to the manufacturer’s instructions. Monocytes were 361 plated in RPMI (Life Technologies) supplemented with 10% SAB (Sigma), 10 mM HEPES (Life 362 Technologies), MEM Non-essential amino acids (Life Technologies) and 25 ng/ml M-CSF (Miltenyi 363 Biotec). Differentiation was obtained after 6 days of culture. Hypoxic cultures were performed in the 364 HypoxyLab chamber authorizing an oxygen partial pressure control (Oxford Optronix, UK).

365 Bone-Marrow Derived Macrophage Differentiation

366 Bone marrow was extracted from the femurs of Balb/c mice in RPMI and then filtered by a 70µm cell 367 strainer. Cells were washed in RPMI and then cultured in RPMI (Gibco) supplemented with 10% of 368 FBS (Gibco) and mice M-CSF at 25 ng/mL (Miltenyi Biotec) for 6 days. Medium was refreshed at day 3. 369 Differentiation was assessed by flow cytometry analysis of surface F4/80 expression.

370 Conditioned medium

371 Macrophages at 1 x 106 cells per well in 12-wells plates were cultured with RPMI supplemented with 372 10% SAB with DMSO (vehicle), Gimeracil 1 µg/ml, 5-FU 0.1 µg/ml or 1 µg/ml or 5-FU 0.1 µg/ml or 1 373 µg/ml with Gimeracil 1 µg/ml during 24h in normoxia and hypoxia. The CM was added to HT-29 and 374 RKO, at 3 x 105 cells per well in 12-wells plates for 48 hours. Then cells were collected, counted and 375 the mortality rate assessed by flow cytometry (Annexin V and 7-AAD).

376 RNAi

377 Small interfering RNA (siRNA) (GE Dharmacon) was transfected at a final concentration of 50 nM using 378 Lipofectamine (RNAiMAx, Life Technologies).

379 Expression of human DPD in mice macrophages

380 RAW264.7 were transduced using lentivirus particles with human DPD (mGFP-tagged) inserted in a 381 pLenti-C-mGFP-P2A-Puro plasmid (Origen Technologies, Rockville, US). Control RAW were obtained 382 using the lentivirus particles containing the plasmid without the DPD ORF sequence (pLenti-C-mGFP- 383 P2A-Puro).

384 RNA isolation and qPCR analysis for gene expression

385 Cells were directly lysed and RNA was extracted using the NucleoSpin RNA kit components (Macherey 386 Nagel) according to the manufacturer’s instructions. Reverse transcription was performed using the 387 iScript Ready-to-use cDNA supermix components (Biorad). qPCR was then performed with the iTaq

388 universal SYBR green supermix components (Biorad) on a CFX96 (Biorad). Quantification was 389 performed using the ΔΔCt method.

390 Immunoblotting

391 Cells were lysed in RIPA buffer supplemented with antiprotease inhibitors (AEBSF 4 mM, Pepstatine A 392 1 mM and Leupeptine 0.4 mM; Sigma Aldrich) and HIF-hydroxylase inhibitor (DMOG 1 mM, Sigma 393 Aldrich). Proteins were quantified by BCA assay (ThermoFischer) and 15 µg of total protein was run on 394 SDS-PAGE gels. Proteins were transferred from SDS-PAGE gels to PVDF membrane (Biorad), blocked 395 with TBS supplemented with 5% milk then incubated with primary antibody at 1µg/mL overnight 4°C. 396 After washing with TBS, the membrane was incubated with a horseradish peroxidase-conjugated 397 secondary antibody (Jackson Immunoresearch). Signal was detected by chemoluminiscence (Chemi- 398 Doc Imaging System, Bio-Rad) after exposition to West Pico ECL (ThermoFischer).

399 Proteomics

400 Cells are directly lysed in Laemmli buffer and prepared and analyzed as previously described (Court et 401 al., 2017). Briefly, the protein equivalent of 300 000 cells for each sample was loaded on NuPAGE Bis- 402 Tris 4–12% acrylamide gels (Life Technologies). Electrophoretic migration was controlled to allow each 403 protein sample to be split into six gel bands. Gels were stained with R-250 Coomassie blue (Bio-Rad) 404 before excising protein bands. Gel slices were washed then dehydrated with 100% acetonitrile (Merck 405 Millipore), incubated with 10mM DTT (Dithiothreitol, Merck Millipore), followed by 55mM 406 iodoacetamide (Merck Millipore) in the dark. Alkylation was stopped by adding 10mM DTT in 25mM 407 ammonium bicarbonate. Proteins were digested overnight at 37 °C with Trypsin/Lys-C Mix (Promega, 408 Charbonnières, France) according to manufacturer's instructions. After extraction, fractions were 409 pooled, dried and stored at −80 °C until further analysis. The dried extracted peptides were 410 resuspendedand analyzed by online nano-LC (Ultimate 3000, Thermo Scientific) directly linked to an 411 impact IITM Hybrid Quadrupole Time of- Flight (QTOF) instrument fitted with a CaptiveSpray ion 412 source (Bruker Daltonics, Bremen, Germany). All data were analyzed using Max- Quant software 413 (version 1.5.2.8) and the Andromeda search engine. The false discovery rate (FDR) was set to 1% for 414 both proteins and peptides, and a minimum length of seven amino acids was set. MaxQuant scores 415 peptide identifications based on a search with an initial permissible mass deviation for the precursor 416 ion of up to 0.07 Da after time-dependent recalibration of the precursor masses. Fragment mass 417 deviation was allowed up to 40 ppm. The Andromeda search engine was used to match MS/MS spectra 418 against the Uniprot human database (https://www.uniprot.org/). Enzyme specificity was set as C- 419 terminal to Arg and Lys, cleavage at proline bonds and a maximum of two missed cleavages are 420 allowed. Carbamidomethylation of cysteine was selected as a fixed modification, whereas N-terminal

421 protein acetylation and methionine oxidation were selected as variable modifications. The “match 422 between runs” feature of MaxQuant was used to transfer identification information to other LC- 423 MS/MS runs based on ion masses and retention times (maximum deviation 0.7 min); this feature was 424 also used in quantification experiments. Quantifications were performed using the label free 425 algorithms. A minimum peptide ratio counts of two and at least one “razor peptide” were required for 426 quantification. The LFQ metric was used to perform relative quantification between proteins identified 427 in different biological conditions, protein intensities were normalized based on the MaxQuant 428 “protein group.txt” output (reflecting a normalized protein quantity deduced from all peptide 429 intensity values). Potential contaminants and reverse proteins were strictly excluded from further 430 analysis. Three analytical replicates from three independent biological samples (donors) were 431 analyzed for each normoxic and hypoxic conditions. Missing values were deduced from a normal 432 distribution (width: 0.3; down shift: 1.8) using the Perseus (version 1.5.5.3) post data acquisition 433 package contained in MaxQuant (www. maxquant.org). Data were further analyzed using JMP 434 software (v.13.0.0, SAS Institute Inc.). Proteins were classed according to the paired Welch test 435 difference (difference between the mean value for triplicate MS/MS analyses for the two conditions 436 compared), and the median fold-change between the two conditions compared.

437 Immunochemistry

438 3 µm thick consecutive tissue sections were prepared from formalin-fixed and paraffin-embedded 439 tissues. Deparaffinization, rehydratation, antigen retrieval and peroxidase blocking were performed 440 on a fully automated system BENCHMARK ULTRA (Roche) according to manufacturer 441 recommendations. The sections were incubated with the following primary antibodies: anti-CD3 clone 442 2GV6 (Roche), anti-CD68 clone Kp1 (Dako) and anti-DPD (ThermoFischer). Revelation was performed 443 using the Ultraview DAB revelation kit (Roche). Nuclei were counterstained with hematoxylin solution 444 (Dako). Images were captured using an APERIO ATS scanner (Leica).

445 Immunofluorescence

446 Formalin fixed and paraffin-embedded human tissue samples were sectioned at 3µm thickness. 447 Samples were de-paraffinized and hydrated by xylene and decreasing concentrations of ethanol baths, 448 respectively. Antigen retrieval was achieved using IHC-TekTM Epitope Retrieval Steamer Set (IW-1102, 449 IHCworld, USA) in IHC-TekTM Epitope Retrieval (IW-1100, IHCworld, USA) for 40 minutes. Nonspecific 450 binding sites were blocked by 1% BSA in PBS. Samples were incubated with the primary antibodies: 451 Monoclonal Mouse Anti-Human CD68 clone PG-M1 at 0.4 µg/mL, Mouse Anti-Human CD163 clone 452 EDHu-1 at 10 µg/mL and DPD polyclonal antibody at 3 µg/mL for 1 hour at room temperature followed 453 by an incubation of secondary antibodies: Alexa Fluor 488 goat anti-mouse IgG (H+L) and Alexa Fluor

454 546 goat anti-rabbit IgG (H+L) both at 4 µg/mL for 30 minutes at room temperature. Nuclei were 455 stained by Hoechst 33342 at 5 µg/mL for 5 minutes at room temperature. Images were captured under 456 20X magnification using ApoTome microscope (Carl Zeiss, Germany) equipped with a camera AxioCam 457 MRm and collected by AxioVision software and analyzed using ImageJ software.

458 Flow cytometry

459 Flow cytometry data was acquired on an Accuri C6 (BD) flow cytometer. The reagents used were: 460 AnnexinV-FITC, mouse anti-F4/80-PE clone REA126 and human anti-CD14-FITC from Miltenyi Biotech 461 and 7-AAD staining solution from BD Pharmingen. Doublet cells were gated out by comparing forward 462 scatter signal height (FSC-H) and area (FSC-A). Dead cells were excluded based on 7AAD or FSC/SSC 463 profile. A least 10,000 events were collected in the analysis gate. Median fluorescence intensity (MFI) 464 was determined using Accuri C6 software (BD).

465 Cancer cell line mRNA DPYD expression

466 Cancer cell line gene expression data were collected from the Cancer Cell Line Encyclopedia (CCLE) 467 (https://portals.broadinstitute.org/ccle). Briefly, sequencing was performed on the Illumina HiSeq 468 2000 or HiSeq 2500 instruments, with sequence coverage of no less than 100 million paired 101 469 nucleotides-long reads per sample. RNaseq reads were aligned to the B37 version of human genome 470 using TopHat version 1.4. Gene and exon-level RPKM values were calculated using pipeline developed 471 for the GTEx project (Barretina et al., 2012; Tsherniak et al., 2017). The cell lines were classified based 472 on their tissue of origin resulting in 24 different groups. The number of cell lines in each group is 473 indicated. The histogram plot was generated using the JMP software (SAS).

474 Mice and human mRNA DPYD expression analysis

475 Genenvestigator 7.5.0 (https://genevestigator.com/gv/) is a search engine that summarizes data sets 476 in metaprofiles. GENEVESTIGATOR® integrates manually curated and quality-controlled gene 477 expression data from public repositories (Hruz et al., 2008). In this study, the Condition Tools Search 478 was used to obtain DPD mRNA levels obtained from human (Homo sapiens) and mice (Mus musculus) 479 in various anatomical parts. Mean of logarithmic level of expression obtained from AFFIMETRIX 480 expression microarrays was used to generate a cell plot from selected results related to macrophages 481 and monocytes. The lowest and the highest expression results in both series (human and mice) were 482 used to evaluate the expression level in the dataset. Cell plot was generated using the JMP software 483 (SAS).

484 DPD activity measurements

485 Since DPD is involved in the hydrogenation of uracil (U) into dihydrouracil (UH2), DPD activity was 486 indirectly evaluated in the cell culture supernatants by determining U and its metabolite UH2 487 concentrations followed by the calculation of the UH2/U ratio. These analyses were performed in the 488 Pharmacology Laboratory of Institut Claudius-Regaud (France) using an HPLC system composed of 489 Alliance 2695 and diode array detector 2996 (Waters), according to a previously described method 490 (Thomas et al., 2016). Uracil (U), Dihydrouracil (UH2), 5-FluoroUracil (5-FU), Ammonium sulfate 99%, 491 Acetonitrile (ACN) gradient chromasolv for HPLC and 2-propanol were purchased from Sigma. Ethyl 492 acetate Scharlau was of HPLC grade and purchased from ICS (Lapeyrouse-Fossat, France). The water 493 used was of Milli-Q Advantage A10 as well as MultiScreen-HV 96-well Plates (Merck Millipore). 494 Calibration ranges were 3.125 – 200 ng/ml for U and 25 - 500 ng/ml for UH2 and 5-FU (5µg/ml) was 495 used as an internal standard. For the 5-FU kinetics experiments, no internal standard was added to 496 the samples and the amount of 5-FU in the supernatant was quantified by the peak area corresponding 497 to 5-FU.

498 Quantification and statistical analysis

499 Statistics were performed using Graph Pad Prism 7 (Graph Pad Software Inc). When two groups were 500 compared, we used a two-tailed student’s t test when variable presented a normal distribution or a 501 Mann-Withney non parametric test otherwise. When more than two groups were compared, we used 502 a one-way ANOVA analysis with Tukey post hoc test. All error bars represent mean with standard error 503 of the mean, all group numbers and explanation of significant values are presented within the figure 504 legends.

505 Data availabilty

506 All MS proteomics data were deposited on the Proteome- Xchange Consortium website 507 (http://proteomecentral.proteomexchange.org) via the PRIDE partner repository, data set identifier: 508 PXD006354.

509 Disclosure of interest: The authors declare no competing interests related to this study.

510 Acknowledgments

511 AM is supported by the ATIP/Avenir Young group leader program (Inserm and La ligue nationale contre 512 le cancer). KG is supported by the ITN (International Training Network) Phys2Biomed project which is 513 funded from the European Union’s Horizon 2020 research and innovation programme under the 514 Marie Skłodowska-Curie grant agreement No 812772.The authors would like to thank Xavier Fonrose, 515 Edwige Col and Floriane Laurent for their technical assistance. The authors thank the zootechnicians 516 of PHTA facility for animal housing and care.

517 Authors’ contributions:

518 AM conceived and designed the study. AM planned and guided the research and wrote the 519 manuscript. MM, MC, KG, and AM performed experiments. MM, MC, KG, MHL, TD, GR, AM analyzed 520 and interpreted data. SM,FT performed DPD activity measurements. MHL performed the 521 histopathological analysis. AM supervised the work carried out in this study and raised funding.

522

523 Supplementary materials

REAGENT or RESOURCE SOURCE IDENTIFIER Antibodies Human CD11b-APC Miltenyi Biotec Cat#:130-110-554 Mouse F4/80-PE clone REA126 Miltenyi Biotec Cat#:130-116-499 Mouse monoclonal anti-human and mice DPYD Clinisciences Cat#:sc-271308 (immunoblots) Rabbit polyclonal anti-human DPYD Life Technologies Cat#:PA522302 (immunochemistry and immunofluorescence) Anti-βactin (immunoblots) Sigma Aldrich Cat#:A2228 Anti-HIF1α (immunoblots) BD Bioscience Cat#:610958 Anti-HIF2α (immunoblots) Clinisciences Cat#:sc-46691 Anti-eIF4E (immunoblots) Clinisciences Cat#:sc-9976 Anti-eIF4E2 (immunoblots) Clinisciences Cat#:GTX82524 Anti-CD68 (immunochemistry) Dako Cat#:M0876 Ant-CD163 (immunochemistry) AbD Serotec Cat#:MCA1853 Anti mouse IgG HRP Life technologies Cat#:G21040 Anti Rabbit IgG HRP Life technologies Cat#:G21234 Anti mouse IgG alexa Fluor 488 Invitrogen Cat#:A11029 Anti rabbit IgG alexa Fluor 546 Invitrogen Cat#:A11010

Biological Samples Human leukoreduction system chambers EFS

Chemicals, Peptides, and Recombinant Proteins 5-Fluorouracil (5-FU) ACCORD 50 mg/mL HEALTHCARE 7-AAD staining solution BD Biosciences Cat#:559925 Actinomycin D Sigma Aldrich Cat#:A9415 AEBSF Sigma Aldrich Cat#:SBR00015-1ML Annexin V-FITC Miltenyi Biotec Cat#:130-093-060 β-mercaptoethanol Sigma Aldrich Cat#:M3148-100ML Bovine Serum Albumin (BSA) solution 30% Sigma Aldrich Cat#:A9576-50ML CD14 microbeads, human Miltenyi Biotec Cat#:130-050-201 Decitabine (5-aza-2’-deoxycitidine) Sigma Aldrich Cat#:A3656-5MG DMOG Sigma Aldrich Cat#:D3695-10MG DMSO for cell culture Dutscher Cat#:702631 DPBS (1X) , no calcium, no magnesium Life Technologies Cat#:14190169 EDTA (0.5M), pH 8.0, RNase-free Life Technologies Cat#:AM9260G Fetal Bovine Serum (FBS), qualified, US Life Technologies Cat#:26140079

17 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.15.341123; this version posted October 15, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

Gimeracil Sigma Aldrich Cat#:SML2075-5MG GSK126 Clinisciences Cat#:HY-13470-5mg HEPES (1M) Life Technologies Cat#:15630056 Histopaque-1077 Sigma Aldrich Cat#:10771-6X100ML Hoescht 33342 Invitrogen Cat#:H3570 Human serum from male AB plasma (SAB) Sigma Aldrich Cat#:H4522-100ML Leupeptin Sigma Aldrich Cat#:L5793-5MG McCoy’s 5A (modified) medium, glutaMAX Life Technologies Cat#:36600088 supplement MEM Non-Essential Amino Acids solution (100X) Life Technologies Cat#:11140035 Opti-MEM I medium, glutaMAX supplement Life Technologies Cat#:51985026 Pepstatin A Sigma Aldrich Cat#:P5318-5MG RPMI 1640 medium, glutaMAX supplement Life Technologies Cat#:61870044 TrypLE Express Enzyme (1X), phenol red Life Technologies Cat#:12605036 Ultrapure DNase/RNase-free distilled water Life Technologies Cat#:10977035

Deposited Data Proteomic data PRIDE PXD006354

Experimental Models: Cell Lines CT26 WT ATCC Cat#:CRL-2638 HT-29 ATCC Cat#:HTB-38 LoVo ATCC Cat#:CCL-229 RAW 264.7 ATCC Cat#:TIB-71 RKO ATCC Cat#:CRL-2577

Experimental Models: Organisms/Strains Female Mice BALB/c Charles River

Oligonucleotides Human B2M forward (housekeeping gene) Eurogentec 5’ – GTGCTCGCGCTACTCTCTC – 3’ Human B2M reverse (housekeeping gene) Eurogentec 5’ – CGGATGGATGAAACCCAGACA – 3’ Human DPYD forward Eurogentec 5’ - CGCAGGACCAGGGGTTTTAT - 3’ Human DPYD reverse Eurogentec 5’ - TGGCAATGGAGAGTGACAGG - 3’ Human HPRT1 forward (housekeeping gene) Eurogentec 5’ – TGCTTTCCTTGGTCAGGCAG – 3’ Human HPRT1 reverse (housekeeping gene) Eurogentec 5’ – TTCGTGGGGTCCTTTTCACC – 3’ Human NDRG1 forward Eurogentec 5’ – GCAGGCGCCTACATCCTAACT – 3’ Human NDRG1 reverse Eurogentec 5’ – GCTTGGGTCCATCCTGAGATCTT – 3’ Human P4HA1 forward Eurogentec 5’ – ACGTCTCCAGGATACCTACAATTT – 3’ Human P4HA1 reverse Eurogentec 5’ – GTCCTCAGCCGTTAGAAAAGATTTG – 3’ Human RPL6 forward (housekeeping gene) Eurogentec 5’ – GTTGGTGGTGACAAGAACGG – 3’ Human RPL6 reverse (housekeeping gene) Eurogentec 5’ – TTTTTGCCGTGGCTCAACAG – 3’ Human SLC2A1 forward Eurogentec 5’ – TGGCCGTGGGAGGAGCAGTG – 3’ Human SLC2A1 reverse Eurogentec 5’ – GCGGTGGACCCATGTCTGGTTG – 3’ Human TBP forward (housekeeping gene) Eurogentec 5’ – GAGAGTTCTGGGATTGTACCG – 3’ Human TBP reverse (housekeeping gene) Eurogentec 5’ – ATCCTCATGATTACCGCAGC – 3’ Human VEGFA forward Eurogentec 5’ – CTTCCTACAGCACAACAAAT – 3’ Human VEGFA reverse Eurogentec 5’ – GTCTTGCTCTATCTTTCTTTG – 3’ Human WASF2 forward (housekeeping gene) Eurogentec 5’ – AAGAAAAGCTGGGGACTTCTG – 3’ Human WASF2 reverse (housekeeping gene) Eurogentec 5’ – GCTACTTGCATCCACGTTTTC – 3’ Mouse b2m forward (housekeeping gene) Eurogentec 5’ – TGGTCTTTCTGGTGCTTGTC – 3’ Mouse b2m reverse (housekeeping gene) Eurogentec 5’ – GTTCAGTATGTTCGGCTTCCC – 3’

Mouse dpyd forward Eurogentec 5’ – TCGCGTGTTCATCGTCTTCA – 3’ Mouse dpyd reverse Eurogentec 5’ – ATAACCTTCCGTGGCGAGAG – 3’ Mouse gusb forward (housekeeping gene) Eurogentec 5’ – GGGACAAAAATCACCCTGCG – 3’ Mouse gusb reverse (housekeeping gene) Eurogentec 5’ – GCGTTGCTCACAAAGGTCAC – 3’ Mouse hprt1 forward (housekeeping gene) Eurogentec 5’ – AGCCCCAAAATGGTTAAGGTTG – 3’ Mouse hprt1 reverse (housekeeping gene) Eurogentec 5’ – ATCCAACAAAGTCTGGCCTGT – 3’ siCTL (Non-targeting pool) Horizon Discovery Cat#:D-001810-10-05 LTD siRNA human DPYD Horizon Discovery Cat#:L-008376-00-0005 LTD siRNA human EIF4E Horizon Discovery Cat#:L-003884-00-0005 LTD siRNA human EIF4E2 Horizon Discovery Cat#:L-019870-01-0005 LTD siRNA human EPAS1 (HIF2a) Horizon Discovery Cat#:L-004814-00-0005 LTD

Recombinant DNA pLenti-C-DPYD-mGFP-P2A-Puro plasmid OriGene Cat#:RC216374L4V pLenti-C-DPYD-mGFP-P2A-Puro plasmid OriGene Cat#:PS100093V

Software and Algorithms Graph Pad Prism 7 GraphPad Software Inc JMP 14 SAS Biorender Biorender

Other iScript Ready-to-use cDNA supermix Biorad Cat#:1708841 iTaq universal SYBR green supermix Biorad Cat#:1725124 HypoxyLab Station Oxford Optronix MicroBCA kit Life Technologies Cat#:23235 MycoAlert kit Lonza Cat#:LT07-118 NucleoSpin RNA (50) Macherey Nagel Cat#:740955.50 Accuri C6 (flow cytometer) BD Biosciences Eclipse TS2 microscope Nikon Transwell Costar Corning Cat#:3460 524

525

526

527

528

529

530

531

532 References

533 Aras, S., and Zaidi, M.R. (2017). TAMeless traitors: macrophages in cancer progression and 534 metastasis. Br. J. Cancer 117, 1583–1591.

535 Barretina, J., Caponigro, G., Stransky, N., Venkatesan, K., Margolin, A.A., Kim, S., Wilson, C.J., Lehár, 536 J., Kryukov, G.V., Sonkin, D., et al. (2012). The Cancer Cell Line Encyclopedia enables predictive 537 modeling of anticancer drug sensitivity. Nature 483, 603–607.

538 Binnewies, M., Roberts, E.W., Kersten, K., Chan, V., Fearon, D.F., Merad, M., Coussens, L.M., 539 Gabrilovich, D.I., Ostrand-Rosenberg, S., Hedrick, C.C., et al. (2018). Understanding the tumor 540 immune microenvironment (TIME) for effective therapy. Nat. Med. 24, 541–550.

541 Court, M., Petre, G., Atifi, M.E., and Millet, A. (2017). Proteomic Signature Reveals Modulation of 542 Human Macrophage Polarization and Functions Under Differing Environmental Oxygen Conditions. 543 Mol. Cell. Proteomics MCP 16, 2153–2168.

544 DeNardo, D.G., and Ruffell, B. (2019). Macrophages as regulators of tumour immunity and 545 immunotherapy. Nat. Rev. Immunol. 19, 369–382.

546 Halbrook, C.J., Pontious, C., Kovalenko, I., Lapienyte, L., Dreyer, S., Lee, H.-J., Thurston, G., Zhang, Y., 547 Lazarus, J., Sajjakulnukit, P., et al. (2019). Macrophage-Released Pyrimidines Inhibit Gemcitabine 548 Therapy in Pancreatic Cancer. Cell Metab. 29, 1390–1399.e6.

549 Henze, A.-T., and Mazzone, M. (2016). The impact of hypoxia on tumor-associated macrophages. J. 550 Clin. Invest. 126, 3672–3679.

551 Ho, J.J.D., Wang, M., Audas, T.E., Kwon, D., Carlsson, S.K., Timpano, S., Evagelou, S.L., Brothers, S., 552 Gonzalgo, M.L., Krieger, J.R., et al. (2016). Systemic Reprogramming of Translation Efficiencies on 553 Oxygen Stimulus. Cell Rep. 14, 1293–1300.

554 Hruz, T., Laule, O., Szabo, G., Wessendorp, F., Bleuler, S., Oertle, L., Widmayer, P., Gruissem, W., and 555 Zimmermann, P. (2008). Genevestigator V3: A Reference Expression Database for the Meta-Analysis 556 of Transcriptomes.

557 Ichikawa, W., Uetake, H., Shirota, Y., Yamada, H., Nishi, N., Nihei, Z., Sugihara, K., and Hirayama, R. 558 (2003). Combination of Dihydropyrimidine Dehydrogenase and Thymidylate Synthase Gene 559 Expressions in Primary Tumors as Predictive Parameters for the Efficacy of Fluoropyrimidine-based 560 Chemotherapy for Metastatic Colorectal Cancer. Clin. Cancer Res. 9, 786–791.

561 Jeong, H., Kim, S., Hong, B.-J., Lee, C.-J., Kim, Y.-E., Bok, S., Oh, J.-M., Gwak, S.-H., Yoo, M.Y., Lee, 562 M.S., et al. (2019). Tumor-Associated Macrophages Enhance Tumor Hypoxia and Aerobic Glycolysis. 563 Cancer Res. 79, 795–806.

564 Kamoshida, S., Shiogama, K., Matsuoka, H., Matsuyama, A., Shimomura, R., Inada, K., Maruta, M., 565 and Tsutsumi, Y. (2003). Immunohistochemical demonstration of dihydropyrimidine dehydrogenase 566 in normal and cancerous tissues. Acta Histochem Cytochem 36, 471–479.

567 Keeley, T.P., and Mann, G.E. (2018). Defining Physiological Normoxia for Improved Translation of Cell 568 Physiology to Animal Models and Humans. Physiol. Rev. 99, 161–234.

569 Marin, J.J.G., Medina, F.S. de, Castaño, B., Bujanda, L., Romero, M.R., Martinez-Augustin, O., Moral- 570 Avila, R.D., and Briz, O. (2012). Chemoprevention, chemotherapy, and chemoresistance in colorectal 571 cancer. Drug Metab. Rev. 44, 148–172.

572 Marusyk, A., Janiszewska, M., and Polyak, K. (2020). Intratumor Heterogeneity: The Rosetta Stone of 573 Therapy Resistance. Cancer Cell 37, 471–484.

574 Mitchem, J.B., Brennan, D.J., Knolhoff, B.L., Belt, B.A., Zhu, Y., Sanford, D.E., Belaygorod, L., 575 Carpenter, D., Collins, L., Piwnica-Worms, D., et al. (2013). Targeting Tumor-Infiltrating Macrophages 576 Decreases Tumor-Initiating Cells, Relieves Immunosuppression, and Improves Chemotherapeutic 577 Responses. Cancer Res. 73, 1128–1141.

578 Murdoch, C., Muthana, M., Coffelt, S.B., and Lewis, C.E. (2008). The role of myeloid cells in the 579 promotion of tumour angiogenesis. Nat. Rev. Cancer 8, 618–631.

580 Olson, O.C., Kim, H., Quail, D.F., Foley, E.A., and Joyce, J.A. (2017). Tumor-Associated Macrophages 581 Suppress the Cytotoxic Activity of Antimitotic Agents. Cell Rep. 19, 101–113.

582 Pinto, M.L., Rios, E., Durães, C., Ribeiro, R., Machado, J.C., Mantovani, A., Barbosa, M.A., Carneiro, F., 583 and Oliveira, M.J. (2019). The Two Faces of Tumor-Associated Macrophages and Their Clinical 584 Significance in Colorectal Cancer. Front. Immunol. 10, 1875.

585 Righi, A., Sarotto, I., Casorzo, L., Cavalchini, S., Frangipane, E., and Risio, M. (2015). Tumour budding 586 is associated with hypoxia at the advancing front of colorectal cancer. Histopathology 66, 982–990.

587 Ruffell, B., and Coussens, L.M. (2015). Macrophages and therapeutic resistance in cancer. Cancer Cell 588 27, 462–472.

589 Ruffell, B., Au, A., Rugo, H.S., Esserman, L.J., Hwang, E.S., and Coussens, L.M. (2012). Leukocyte 590 composition of human breast cancer. Proc. Natl. Acad. Sci. 109, 2796–2801.

591 Salonga, D., Danenberg, K.D., Johnson, M., Metzger, R., Groshen, S., Tsao-Wei, D.D., Lenz, H.J., 592 Leichman, C.G., Leichman, L., Diasio, R.B., et al. (2000). Colorectal tumors responding to 5- 593 fluorouracil have low gene expression levels of dihydropyrimidine dehydrogenase, thymidylate 594 synthase, and thymidine phosphorylase. Clin. Cancer Res. Off. J. Am. Assoc. Cancer Res. 6, 1322– 595 1327.

596 Shirota, Y., Ichikawa, W., Uetake, H., Yamada, H., Nihei, Z., and Sugihara, K. (2002). Intratumoral 597 dihydropyrimidine dehydrogenase messenger RNA level reflects tumor progression in human 598 colorectal cancer. Ann. Surg. Oncol. 9, 599–603.

599 Shree, T., Olson, O.C., Elie, B.T., Kester, J.C., Garfall, A.L., Simpson, K., Bell-McGuinn, K.M., Zabor, 600 E.C., Brogi, E., and Joyce, J.A. (2011). Macrophages and cathepsin proteases blunt chemotherapeutic 601 response in breast cancer. Genes Dev. 25, 2465–2479.

602 Sica, A., and Mantovani, A. (2012). Macrophage plasticity and polarization: in vivo veritas. J. Clin. 603 Invest. 122, 787–795.

604 Siegel, R.L., Miller, K.D., and Jemal, A. (2020). Cancer statistics, 2020. CA. Cancer J. Clin. 70, 7–30.

605 Soong, R., Shah, N., Salto-Tellez, M., Tai, B.C., Soo, R.A., Han, H.C., Ng, S.S., Tan, W.L., Zeps, N., 606 Joseph, D., et al. (2008). Prognostic significance of thymidylate synthase, dihydropyrimidine

607 dehydrogenase and thymidine phosphorylase protein expression in colorectal cancer patients 608 treated with or without 5-fluorouracil-based chemotherapy. Ann. Oncol. 19, 915–919.

609 Tanaka-Nozaki, M., Onda, M., Tanaka, N., and Kato, S. (2001). Variations in 5-Fluorouracil 610 Concentrations of Colorectal Tissues as Compared with Dihydropyrimidine Dehydrogenase (DPD) 611 Enzyme Activities and DPD Messenger RNA Levels. Clin. Cancer Res. 7, 2783–2787.

612 Thomas, F., Hennebelle, I., Delmas, C., Lochon, I., Dhelens, C., Tixidre, C.G., Bonadona, A., Penel, N., 613 Goncalves, A., Delord, J.P., et al. (2016). Genotyping of a family with a novel deleterious DPYD 614 mutation supports the pretherapeutic screening of DPD deficiency with dihydrouracil/uracil ratio. 615 Clin. Pharmacol. Ther. 99, 235–242.

616 Tsherniak, A., Vazquez, F., Montgomery, P.G., Weir, B.A., Kryukov, G., Cowley, G.S., Gill, S., 617 Harrington, W.F., Pantel, S., Krill-Burger, J.M., et al. (2017). Defining a Cancer Dependency Map. Cell 618 170, 564–576.e16.

619 Uniacke, J., Holterman, C.E., Lachance, G., Franovic, A., Jacob, M.D., Fabian, M.R., Payette, J., Holcik, 620 M., Pause, A., and Lee, S. (2012). An oxygen-regulated switch in the protein synthesis machinery. 621 Nature 486, 126–129.

622 Vasan, N., Baselga, J., and Hyman, D.M. (2019). A view on drug resistance in cancer. Nature 575, 623 299–309.

624 Wilson, Z.E., Rostami-Hodjegan, A., Burn, J.L., Tooley, A., Boyle, J., Ellis, S.W., and Tucker, G.T. (2003). 625 Inter-individual variability in levels of human microsomal protein and hepatocellularity per gram of 626 liver. Br. J. Clin. Pharmacol. 56, 433–440.

627 Wu, R., Nie, Q., Tapper, E.E., Jerde, C.R., Dunlap, G.S., Shrestha, S., Elraiyah, T.A., Offer, S.M., and 628 Diasio, R.B. (2016). Histone H3K27 Trimethylation Modulates 5-Fluorouracil Resistance by Inhibiting 629 PU.1 Binding to the DPYD Promoter. Cancer Res. 76, 6362–6373.

630 Ye, L., Zhang, T., Kang, Z., Guo, G., Sun, Y., Lin, K., Huang, Q., Shi, X., Ni, Z., Ding, N., et al. (2019). 631 Tumor-Infiltrating Immune Cells Act as a Marker for Prognosis in Colorectal Cancer. Front. Immunol. 632 10, 2368.

633 Yin, Y., Yao, S., Hu, Y., Feng, Y., Li, M., Bian, Z., Zhang, J., Qin, Y., Qi, X., Zhou, L., et al. (2017). The 634 Immune-microenvironment Confers Chemoresistance of Colorectal Cancer through Macrophage- 635 Derived IL6. Clin. Cancer Res. Off. J. Am. Assoc. Cancer Res. 23, 7375–7387.

636 Zhang, X., Chen, Y., Hao, L., Hou, A., Chen, X., Li, Y., Wang, R., Luo, P., Ruan, Z., Ou, J., et al. (2016). 637 Macrophages induce resistance to 5-fluorouracil chemotherapy in colorectal cancer through the 638 release of putrescine. Cancer Lett. 381, 305–313.

639 Zheng, J.-F., and Wang, H.-D. (2005). 5-Fluorouracil concentration in blood, liver and tumor tissues 640 and apoptosis of tumor cells after preoperative oral 5’-deoxy-5-fluorouridine in patients with 641 hepatocellular carcinoma. World J. Gastroenterol. 11, 3944–3947.

642

643

644

645 Figure Legends

646 Figure 1 Macrophages confer a chemoresistance to 5-FU in hypoxia

647 (A) Growth inhibition of HT-29 cells by CM(Ø), 5-FU and CM(5-FU) in normoxia (blue) and in hypoxia 648 (red), 5-FU was used at 1µg/mL (n=3).

649 (B) Growth inhibition of RKO cells by CM(Ø), 5-FU and CM(5-FU) in normoxia (blue) and in hypoxia 650 (red) 5-FU was used at 0.1µg/mL and 1µg/mL (n=3).

651 (C) Protein heat map of macrophages in hypoxia and normoxia. Proteins were selected by a fold 652 change >2 and p-value < 0.01. Proteins were organized according to descending mean z-score of 653 hypoxic proteins.

654 (D) Schematic presentation of the rate-limiting steps of the pyrimidine degradation pathway involving 655 DPD.

656 (E) Chemical structures of uracil and 5-fluorouracil

657 (F) Immunoblot analysis of DPD expression in human macrophages differentiated in normoxia and 658 hypoxia (n= 10).

659 (G) dihydrouracil/uracil ratio measured by HPLC in the macrophage supernatant from macrophages 660 cultured in normoxia and hypoxia (n=3).

661 Error bars represent mean ± sem, *p<0.05, **p<0.01, *** p<0.001, ****p<0.0001, ns=non significant

662

663 Figure 2 Chemoresistance to 5-FU is driven by DPD activity in macrophages

664 (A) Kinetics of 5-FU degradation by macrophages obtained by HPLC. 5-FU initial concentration was 665 1µg/mL. Normoxia appears in blue and hypoxia in red. 5-FU without macrophages was stable during 666 the 48h period of study in normoxia (full black circle) and in hypoxia (empty black circle) (n=3).

667 (B) Chemical structure of gimeracil a specific inhibitor of DPD.

668 (C) 5-FU degradation due to DPD activity in macrophages inhibited by gimeracil. 5-FU initial 669 concentration was 1 µg/mL. Normoxia appears in blue and hypoxia in red (n=3).

670 (D) Induction of death in HT-29 in normoxia (blue) and in hypoxia (red) by non-conditioned medium 671 (empty histogram) and conditioned medium (full histogram). Dead cells were defined as AnnexinV+ 672 cells in flow cytometry. 5-FU was used at 1 µg/mL and gimeracil at 1 µg/mL (n=4).

673 (E) Inhibition of growth and death induction in 3D tumoroïd of HT-29 cells in normoxia and hypoxia. 674 Tumoroïds were exposed to CM(vehicle), CM(5-FU 1 µg/mL) and CM(5-FU 1 µg/mL + gimeracil 1 675 µg/mL). Picture were obtained with a phase contrast microscope, scale bar is 200 µm (n=8).

676 (F) Induction of death in RKO in normoxia (blue) and in hypoxia (red) by non-conditioned medium 677 (empty histogram) and conditioned medium (full histogram). Dead cells were defined as AnnexinV+ 678 cells in flow cytometry. 5-FU was used at 1 µg/mL and gimeracil at 1 µg/mL (n=4).

679 (G) Immunoblot of DPD expression in macrophages exposed to 5-FU at 50 µg/mL during 48 h (n=3).

680 (H) Immunoblot of DPD expression in macrophages transwell co-cultured with HT-29 or RKO in 681 normoxia and hypoxia (n=3).

682 (I) Induction of death in RKO cells directly co-cultured with macrophages in normoxia (blue) and 683 hypoxia (red). 5-FU was used at 1 µg/mL and gimeracil at 1 µg/mL. Dead cells were defined as CD11b- 684 AnnexinV+ cells in flow cytometry. Gating strategy is represented on left panel. Dead cell 685 quantification is represented on the right panel (n=4).

686 Error bars represent mean ± sem, *p<0.05, **p<0.01, *** p<0.001, ****p<0.0001, ns=non significant

687

688 Figure 3 A translational mechanism controls DPD expression involving the eIF4FHypoxic complex

689 (A) Immunoblot of DPD expression in normoxia (N) and hypoxia (H) and during transition from 690 normoxia to hypoxia (NH) and hypoxia to normoxia (HN) (n=4).

691 (B) Quantification of the expression of DPD from immunoblots of figure 3A for NH and HN transitions 692 (n=4).

693 (C) Mathematical model fitting DPD expression curves from Figure 3B (left panel). DPD half-life 694 extracted from the model (right panel).

695 (D) DPD activity determined in the supernatant of macrophages cultured in normoxia and hypoxia for 696 24h. Macrophages were previously differentiated in normoxia and hypoxia permitting to study the 697 activity of DPD following transitions (n=3).

698 (E) Immunoblot of DPD expression in profound hypoxia for 24h and 48h (PO2 =7mmHg; n=4).

699 (F) Immunoblot analysis of DPD and HIF-1α in macrophages exposed to CoCl2 250 µM for 4h. β-actin 700 is used as control loading (n=3).

701 (G) mRNA expression ratio NH/N and HN/H transitions determined by qPCR for the following genes 702 DPYD, VEGF-A, NDRG1, P4HA1 and SLC2A1. Macrophages were previously cultured in normoxia or 703 hypoxia (n=3).

704 (H) Immunoblot of DPD expression during NH (hypoxia PO2 =7mmHg) transition with macrophages 705 exposed previously to Actinomycin D 1 µg/mL 20 minutes (n=3).

706 (I) 3’UTR sequence of the DPYD gene containing a CGG RBM4 binding motif highlighted in red 707 (Genecode transcript variant 1 ENST00000370192.7 https://genome.ucsc.edu/).

708 (J) Immunoblot of DPD expression during NH transition under siRNA silencing of eIF4E2 and eIF4E 709 (n=3).

710 (K) Immunoblot of DPD expression during NH transition under siRNA silencing of HIF2α (n=3).

711

712 Figure 4 Macrophages harbor the main DPD expression in colorectal cancer

713 (A) RNAseq analysis of DPYD expression in various cancer cell lines from the Cancer Cell Line 714 Encyclopedia (CCLE). Colon cancer cell lines are in red.

715 (B) Immunochemistry analysis of CD68 (upper panel) and DPD (lower panel) expression in liver 716 metastasis of colorectal cancer (n=15; scale bar = 200 µm).

717 (C) Immunochemistry analysis of DPD expression in various cell populations in liver metastasis. Red 718 arrowheads highlight macrophages, black arrows point to metastatic cancerous cells (n=15; scale bar 719 = 60 µm).

720 (D) Immunochemistry analysis of DPD expression in primary tumors. Red arrowheads are 721 macrophages, black arrows point to cancer cells and the black arrowhead identifies a tripolar mitosis 722 of a cancer cell (n=15; scale bar = 60 µm).

723 (E) Immunofluorescence staining in liver metastatic tissues. CD68 is in green, DPD in red, nuclei are 724 stained by Hoescht in blue (n=4; scale bar= 50 µm).

725 (F) Immunofluorescence staining in primary tumors. CD68 is in green, DPD in red, nuclei are stained 726 by Hoescht in blue (n=4; scale bar= 50 µm).

727

728 Figure 5 Rodents’ macrophages do not express DPD and transduced human DPD in mice 729 macrophages leads to 5-FU chemoresistance in vivo

730 (A) BMDM were differentiated in normoxia and hypoxia. F4/80 was studied by flow cytometry (left 731 panel) and mDPD expression by immunoblot (n=5). Mice liver was used as a positive control for DPD.

732 (B) RAW264.7 macrophages were cultivated in normoxia and in hypoxia. mDPD expression was 733 studied by immunoblot. No production of dihydrouracil was found in RAW supernatant using HPLC 734 (nd=non detected).

735 (C) Microarray analysis of DPYD mRNA expression in monocytes/macrophage populations in mice and 736 humans. In each group the highest and lowest level of expression was used to scale the heat map.

737 (D) mDPYD mRNA level of expression in RAW macrophages exposed to decitabine at 5µM during 24h 738 (n=3).

739 (E) Transduced GFP and hDPD-GFP proteins levels of expression analyzed by flow cytometry.

740 (F) Immunoblot of RAW macrophages transduced to express GFP and hDPD-GFP

741 (G) Growth inhibition of CT-26, after 48h, under the presence of CM containing 5-FU 0.1µg/mL 742 exposed to macrophages WT, expressing GFP or hDPD-GFP for 24h. Gimeracil was used to block DPD 743 activity at 1 µg/mL.

744 (H) Tumor assay was performed on female Balb/c mice of 7 weeks. 106 CT-26 and 106 RAW were 745 implanted subcutaneously. After ten days daily bolus of 5-FU 25 mg/kg were injected intraperitoneally 746 according to the timeline represented.

747 (I) Tumors growth was followed during the protocol (n=7 in each group).

748 (J) Tumors weight was determined at day 24 (n=7 in each group).

749

750 Figure 6 DPD expression in macrophages confers a chemoresistance to 5-FU under the control of 751 oxygen

752 Schematic of the chemoresistance mechanism due to DPD expression in hypoxic macrophages under 753 the control of the eIF4FHypoxic complex. 5-FU degradation is mainly performed by macrophages in the 754 tumor microenvironement providing a protection for cancer cells, free to proliferate.

755

756

757

758 Supplemental Figures

759 Figure S1

760 (A) Immunoblot analysis of DPD expression in human macrophages, RKO and HT-29 cells in normoxia 761 and hypoxia (n=3).

762 (B) Dihydrouracil to uracil ratio in the supernatant of human macrophages, RKO and HT-29 cells in 763 normoxia and hypoxia measured by HPLC (n=3).

764 (C) qPCR analysis of DPYD mRNA in human macrophages, RKO and HT-29 cells (n=3). nd=non detected

765 (D) qPCR analysis of DPYD mRNA in RKO cells exposed to GSK-126 (an EZH2 inhibitor) at 5 µM during 766 96h (n=3)

767 Figure S2

768 (A) Immunofluorescence staining in liver metastatic tissues. CD163 is in green, DPD in red, nuclei are 769 stained by Hoescht in blue (n=4; scale bar= 50 µm).

770 (B) Immunofluorescence staining in primary tumors. CD163 is in green, DPD in red, nuclei are stained 771 by Hoescht in blue (n=4; scale bar= 50 µm).

772 (C) Immunoblot analysis of DPD expression in human macrophages exposed to IL-4 (20 ng/mL), IL-13 773 (20 ng/mL), IL-10 (25 ng/mL), IFNγ (10 ng/mL), IL-6 (20 ng/mL) and GM-CSF (25ng/mL) compared to 774 hypoxic condition for 24h (n=3). β-actin is used as control loading (n=3).

775

776

Figure 1

HT-29 Hypoxia Normoxia A C 1.7 DPD *** H T 2 9 ns NDRG1 1.4 FTH1 1 PFKP 80 ** * 0.6 SLC25A19 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.15.341123; this versionN o r mpostedNDUFS7o x ia October 15, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder.0.3 All rightsRNF40 reserved. No reuse allowed without permission. CLEC5A -0.1 H y p o x ia

n TRAPPC5 606 0 -0.5

o C4BPA i

t -1 NSDUFS8 i CNOT7

b -1.4 i

inhibition HK2

h 40 -1.9 MRPL2 n

i -2.3

SP100

f ENO2 o

DOCK5

% 2 0 NQO1

growth 20 NDUFS3

% FGG SL43A2 0 SNX8 0 C1QB CTSC TTR LSM2 SDSL TCEB1 MKKS STX8 FAM49A ATOX1 ZNF185 CSNK2A1 B R KRKOO M C RAP2A ns CST3 CYB5A 1001 0 0 **** COMMD6 N o r mCREG1o x ia ATP5L 88 00 **** H y p oUAP1L1x ia

n HLA-DRA o

i FGD2 t 606 0 **** i ABCD3

b COPS8 i inhibition

h EIF3K 44 00

n VPS45 i

COMMD5 f

o 2 0 UCHL3 20 NAAA

% SULT1A1 growth 00 NRAS % PPP1CC ALDH1A --202 0 FDX1 GSN HLA-DRB3 HLA-DRB1

D DPD E Uracil Uracil-H2 NADPH + H+ NADP+ DPYS UPB1 uracil 5-fluorouracil

β-alanine

J 6 F G 2002 0 0 **

1501 5 0

U /

2 1 0 0 DPD UH /U 100H

2 U 505 0 β-actin 00 N H 5 8 108 0 0 0 0 0 Figure 2 5 F U G i m e r a c i l 6 C 1 0 0 100 0 0 0 A 5 B * *** 6 6100 0 0 0 0 8 0 0 0 0 50

) 8 10

A ( 5 n 6 0 0 0 0 0

o 6 10

t a

r 5 t 4 0 0 0 0 0 4 n 10 * 5 (AU)

e 44 0 100 0 0 0

n o

(AU) c * 5

U 22 0 100 0 0 0

F 5 FU concentration FU concentration

5 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.15.341123; this version - posted October 15, 2020. The copyright holder for this 2 2100 0 0 0 0 preprint*** (which was not certified by peer review) isg imeracilthe author/funder. All rights5 reserved.00 No reuse allowed without permission. FU concentration FU concentration

- a i l i l i i a 5 x c c a x a o r o r m e p e r m y m o i H i N G G * + +

0 0 i a i a 0 2 4 4 8 x x o o 0 24 48 p m T i m e ( h ) r y o H Time (h) N

D N = 4 M CHT d -e29a d H T - 2 9 E ** 252 5 **** Veh 5FU 5FU + Gimeracil ns ***

2 0

s 20

l l

e N o r m o x ie

1 5 Normoxia

cells d

a N o r m o x ie a v e c M 0

e d

1 0 H y p o x ie f 10

dead

H y p o x ie a v e c M 0 % % 5 Hypoxia

l l h m e m i / / m V g G g µ µ 5FU 50µg/mL G 1 1 CTL U m 48h F i 5 G + l N H N H / m g N o r m o x ia µ 1 M a c r o N o r m o x ia U DPD F 5 H y p o x ia F C o n d i tRKOi o n s M a c r o H y p o x iaβ-actin R****K O M C c o m b i n é s % ****f i g u r e 2 5500 ns **** H

4 0 s

4l 0

l Normoxia Hypoxia e

cells c 3 0

30d a e Co-culture Ø

d Ø

2 0 f

dead 20

% % 101 0 DPD

00 β-actin L 1 1 1 T m C U i i m F C G G 5 M + + + C C 1 M M U F 5 + C M Macrophages + RKO R K O C o C u l t u r e C A L L C E L A 5 + f i g u r e 2 **** 606 0 **** ns **** N o r m o x ia

H y p o x ia

I s

l l

e 4 0

c 40

macrophages cells

f dead

o 202 0

CD11b % 7AAD RKO %

FSC 00

1 h m m e i i v U g g F 0 0 + Annexin V 5 M 1 M 0 U M O O F K K O 5 R R K 0 t l t l R M u t u c l c u O o c o K c c o R c l t u c o c Figure 3

Normoxia to Hypoxia (3% O ) D6 D7 D8 D10 B 2 A Hypoxia (3% O2) to Normoxia N N NH N NH N NH c i n e t i q u e DPD 6 0 0 060

N H

t i

n 4 0 0 0 0

β-actin units 4 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.15.341123y ; this version posted October 15, 2020. The copyright holder for this

r H N a

preprint (which was not certified by peer review) is the author/funder.r All rights reserved. No reuse allowed without permission. t H H HN h HN H HN i b 2 0 0 0 0 r 2 DPD A Arbitrary 00 β-actin J6 00 11 22 33 44 55 t Time (d) C 푑 퐷푃퐷 = 푘 푂 − 푘′ 푂 퐷푃퐷 Half life N to H 0.88 day +/- 0.22 푑푡 2 2 Synthesis term Degradation term Half life H to N 1.57 days +/- 0.53

D J 8 E F CoCl2 250 µM 4h 8800 24h 48h - + Normoxia + - + - 606 0 * DPD U Hypoxia <1% O - + - + /U

/ 2 2 2 4 0

4H 0 ** U DPD UH 202 0 HIF1α β-actin 00 N H N H β-actin N NHN HNH H G H Normoxia L oHypoxiag 2 N H / N a n6hd H NHypoxia/ H p a p i e r 6 h Normoxia 6h

66 n

o 4 i 4

s Normoxia Hypoxia 6h

e r

mRNA p

x 2

e 2 Actinomycin D - + - +

n e

g 0

0 DPD

a l

e - 2

-R 2 ratio H/N ratio β-actin 2 -4- 4 og l DPYD VEGF-A NDRG1 P4HA1 SLC2A1

I DPYD 3’UTR C o n d i t i o n s D P Y D 3’

ggt tgcacatacgggctctgacV E G F A aaa

N D R G 1

1 771 P 4 H A 1 789 1233 J S L C 2 A 1 K Norm. Hypoxia 24h Normoxia Hypoxia 24h

DPD DPD eIF4E HIF2α eIF4E2 β-actin 4ns8 h 1.51 . 5 β-actinT r a n s f e c t i o n D P Y D N r o m a l i z e d a c t i n e ** N C T L

N S i H I F 2 a 11.5. 5 * * ns 11. 0 N H C T L actin -

N H S i H I F 2 a n β

i ns t 1 . 0 c 1

actin A

- 0 . 5

- 0.5

b / β

DPD/

D Y 0 . 5 0.5P

D 0 . 0

DPD/ 0

L a L a T 2 T 2 F F C I C I N H H H 0 . 0 i i 0 N S S N H H H N H N L N N N T L E 2 C T 4 E C F 4 I F E I E Figure 4 A B 66 CD68

) 44

bioRxiv preprint doi: https://doi.org/10.1101/2020.10.15.341123; this version posted October 15, 2020. The copyright holder for this M DPYD 2 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

K 2

2 g

o 0 l 0 RPKM ( RPKM 2 og l --22 DPD

--44

) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) 1 6 1 2 1 2 1 6 6 2 8 7 8 2 6 6 5 7 7 8 8 7 1 9 ( ( ( ( 6 1 1 3 4 2 1 3 2 2 9 7 5 2 4 5 2 3 3 5 ( ( ( ( ( ( ( ( ( ( ( ( ( E ( X ( ( T 1 1 ( ( E I ( ( T D A T S T A Y E S C N R Y T H E E N M I I M I I V E R C A C E N U G E S A C U N E L A R U I T O R K I E N U A T S T U A A N O G R V A A T S E G S I R S E R E R R R D T U S S E S B A L V M I S Y C N I S T L T T _ L R O T Y C H I O T E A K E O H _ _ Y S P N P T B R _ T N T T I E Y G _ M T _ A R P N _ _ O S I V R F S I P A C O L C S I I _ U T A I D O L E L T E I S O S N M E N A I O G B I V E E R O M O R R G N S I U P A E O L D O N T O T _ U L R A A A E M R A E _ T A R N H E E P C P

U C D DPD DPD

E CD68 DPD Hoescht Merged etastasis m Liver

F CD68 DPD Hoescht Merged tumor Primary Figure 5 A B Mice Mice RAW RAW Liver Liver N H Normoxia + - + - + - + - Hypoxia - + - + - + - + mDPD

events bioRxiv preprint doi: https://doi.org/10.1101/2020.10.15.341123; this version posted October 15, 2020. The copyright holder for this # preprint (which wasmDPD not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. UH2 nd nd β-actin β-actin F4/80 C D 16.5 hDPYD mDPYD 16.7 mDPYD 15.9 15.2

15.3 n o

Bone marrow 13.7 i 1 0 , 0 0 04 14.7 s 10

monocyte Liver 12.2 s * e 14.1 r 3 10.7 p 88 ,100 0 0

Peripheral blood x e

13.5 expression Colon 9.2 A 3 12.1 monocyte 6 , 0 0 0 8.5 N 6 10 10.7 R Bone marrow m Monocyte derived 3 7.9 4 , 0 0 0

mRNA 4 10 9.4 D macrophage mononuclear cell7.3 Y 8 P 3 D 2 , 0 0 0

Peritoneal 6.6 2 10 e

6.6 Liver v i

macrophage 6 t

a 0

l 0

Relative Relative e

Bone marrow r V e h i c l e D e c i t a b i n e Alveolar macrophage macrophage Peripheral blood Bone marrow derived mononuclear cell macrophage

Colon Bone marrow monocyte Tonsillar memory B cell Bone marrow haematopoeitic SC E F G I n h i b i t i o n C T - 2 6 M C R A W RAW GFP RAW DPD-GFP 1001 0 0

8800 WT WT n * * *

DPD-GFP o

i t i 6 0 (~140 kDa) b 60

Transduced Transduced i h inhibition

number

n i 4 0

f 40

Cell ***

β-actin % 202 0 growth

GFP % 00 / / / L g g g / i m m µ µ µ D g / 1 T 1 P 1 G g . . F . Y µ 0 0 0 P 1 µ W G . + 1 D . U C U U 0 D 0 C F F F C U Y 5 M 5 M 5 H U M F P L F L 5 D 5-FU IP 5-FU IP L 5 m m m C injection 25mg/kg/d injection 25mg/kg/d M SC injection L m CT-26 + RAW sacrifice

D0 D10 D17 D24 I c u r v e s a n a l y s i s * J 0.30 . 3

10001 0 0 0 )

) G F P g

) 3 * ** ( 3

0 . 2 m

8008 0 0 D P Y D s 0.2

a (

CT-26 + RAW hDPD-GFP

mass (g) mass

6006 0 0 r

m CT-26 + RAW GFP o

u 0 . 1 l

m 0.1

u t

v 4 0 0

volume volume (mm 400

Tumor

u o 2 0 0 0 0. 0

m 200 u

Tumor P D t F Y G P 0 D 5 101 0 151 5 202 0 252 5 Time postt i m implantatione ( d a y s ) (days) Figure 6

R K O H T 2 9 v s M a c r o Mφ RKO HT29 A B 8 0

bioRxivN preprintH doi:N https://doi.org/10.1101/2020.10.15.341123H N H ; this version posted October* 15, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. 606 0

DPD /U

U /

2 4 0

UH 40

H U

Control 202 0 loading

N N HH N N H H N N HH o o O O 9 9 r r 2 2 c c K K T T a a R R H H M MM φ RKO HT29

DPYD DPYD C D P Y D D e x p r e s s i o n w i t h G S***K 1 2 6 o n R K O 9 6 h f i g u r e p a p i e r

q P C R 2 8 / 0 2 / 2 0 R e l a t i v e D P Y D m R N A e x p r e s s i o n o

i s 4 0

40 s

r p

n 4

o x i 4

3 0 s

expression

e s

30 e

expression

A 2 0

P N

20R

mRNA 2

e 2

e v

mRNA

a l

1 0 i e

10 t

a l

nd e 0 0R 0 0 Relative 9 O s 2 - K e Relative Relative T R g H a h h M p e o µ r c V a 5 m y 6 r a 2 i m 1 r P K S C e l l s G

C o n d i t i o n s Supplemental Figure 2

A CD163 DPD nucleus merged

B CD163 DPD nucleus merged tumor Primary

Stimulation 48h - IFNγ IL4 IL6 IL10 IL13 GM-CSF DPD β-actin bioRxiv preprint doi: https://doi.org/10.1101/2020.10.15.341123; this version posted October 15, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

synchronous metastasis: 1 Desease free metachronous Nomber of liver Targeted decease from survival Total number of Sex Age Comorbidities metastasis: 2 metastasis First line chemotherapy therapy cancer (months) treatment lines F 62 / 1 1 Fluorouracil + Irinotecan anti--VEGF yes 25 1 F 65 diabetes, Hypertension 2 2 Fluorouracil + Irinotecan none yes 73 3 M 61 / 1 2 Fluorouracil + Oxaliplatin anti-EGFR yes 63 3 M 64 Cardiopathy, Hypertension, diabetes 1 3 Fluorouracil + Oxaliplatin anti--VEGF alive 77 1 M 74 Hypertension, asthma 1 1 Fluorouracil + Oxaliplatin none alive 25 1 M 78 / 2 1 Fluorouracil + Oxaliplatin none alive 56 1 M 69 / 2 2 Fluorouracil + Oxaliplatin anti--VEGF alive 43 4 M 45 alcoolism, hepatic cirrhosis 1 > 5 Fluorouracil + Oxaliplatin none yes 53 4 M 73 Asthma 2 1 Fluorouracil + Oxaliplatin none alive 142 1 F 48 / 1 2 Fluorouracil + Oxaliplatin anti--VEGF yes 66 5 F 45 / 1 2 Fluorouracil + Oxaliplatin none yes 35 2 M 57 / 1 > 5 Fluorouracil + Irinotecan + Oxaliplatin none alive 42 1 M 72 cardiac insufficinecy, diabetes, chronic obstrutive pneumopathy 2 5 Fluorouracil + Irinotecan + Oxaliplatin anti--VEGF yes 76 5 F 76 / 2 1 Fluorouracil none yes 100 2 M 66 Alccolism 1 2 Fluorouracil + Oxaliplatin none yes 37 2