Neutrophils are the most abundant circulating leucocytes and are essential for innate immunity. In cancer, pro- or antitumor properties have been attributed to tumor-associated neutrophils (TAN). Here, focusing on TAN accumulation within lung tumors, we identify GLUT1 as an essential glucose transporter for their tumor supportive behavior. Compared with normal neutrophils, GLUT1 and glucose metabolism increased in TANs from a mouse model of lung adenocarcinoma. To elucidate the impact of glucose uptake on TANs, we used a strategy with two recombinases, dissociating tumor initiation from neutrophil-specific Glut1 deletion. Loss of GLUT1 accelerated neutrophil turnover in tumors and reduced a subset of TANs expressing SiglecF. In the absence of GLUT1 expression by TANs, tumor growth was diminished and the efficacy of radiotherapy was augmented. Our results demonstrate the importance of GLUT1 in TANs, which may affect their pro- versus antitumor behavior. These results also suggest targeting metabolic vulnerabilities to favor antitumor neutrophils.
Lung tumor support and radiotherapy resistance depend on GLUT1-mediated glucose uptake in tumor-associated neutrophils, indicating that metabolic vulnerabilities should be considered to target both tumor cells as well as innate immune cells.
Neutrophils are critical for the innate immune response (1). The expected short half-life of these cells might have contributed to underestimating their importance in tumor progression in past research. However, neutrophil functions in tumors have begun to be revealed in different types of cancer including breast and lung cancers (2). Although they have now been considered in several cancer models, only a few studies used autochthonous mouse models closely recapitulating the characteristics of the human disease (3). Neutrophils with protumor properties have been extensively described in the context of breast cancer metastasis (4, 5), and have also been linked to immune-suppression and pro-angiogenesis (6, 7). Immune signature analyses of lung adenocarcinoma (LUAD) from the KrasLox-STOP-Lox-G12D/WT; Tp53Flox/Flox (KP) mouse model (8) demonstrated that neutrophils are the most prevalent immune cells within the tumor mass (9) and may possess tumor-supportive properties (10). Specifically, a TAN subset defined by high SiglecF membrane protein expression showed a tumor-associated transcriptional profile and an increased reactive oxygen species production in a KP cell-derived transplantation model in vivo. Although this subset remains to be fully characterized, SiglecFhigh TANs have tumor-promoting functions compared with SiglecFlow cells (11). These suggest that gaining fundamental knowledge about neutrophil heterogeneity in cancer will enable selective targeting of protumor neutrophils without affecting other neutrophils that are essential for the host defense.
Although multiple metabolic alterations of tumor cells are known to help them survive and proliferate in harsh environmental conditions (12, 13), much less is known about the rewired metabolism of nontumor cells from the tumor microenvironment. Because neutrophils are exposed to the same hypoxic environment with possibly reduced access to nutrients, we hypothesized that their glucose metabolism changes within solid tumors and impacts tumor development. To test this, we explored the effect of the lung tumor microenvironment on TAN metabolism, survival, and turnover as well as the consequences of glucose transporter GLUT1-deficient (Glut1KO) TANs on disease progression.
Materials and Methods
A Key Resources Table is provided as Supplementary Table S1.
KrasLox-STOP-Lox-G12D/WT; Tp53Flox/Flox (KP) (RRID:IMSR_JAX: 008179, RRID:IMSR_JAX:008462) mice were described before (10) and were used in this study for the experiments comparing wild-type TANs to wild-type HLNs, represented in Fig. 2. KrasFrt-STOP-Frt-G12D/WT (KFrt) (RRID:IMSR_JAX:008653) mice, previously generated by T. Jacks (14), Massachusetts Institute of Technology, were bred with Tp53Frt/Frt (PFrt) (RRID:IMSR_JAX: 017767) mice (15). The resulting animals, KrasFrt-STOP-Frt-G12D/WT; Tp53Frt/Frt (KPFrt) mice, were kindly provided by D.G. Kirsch, Duke University Medical Center in a mixed 129-C57BL/6 background. The generation of the Glut1Flox/Flox (G1) (RRID:IMSR_JAX:031871) mice was described previously (16), and the mice in a C57BL/6 background were provided by E. Dale Abel, University of Iowa. Glut3Flox/Flox (G3) mice were produced at the Duke University Transgenic Facility upon blastocyst microinjection of targeted ES cells obtained from KOMP Repository (project CSD48048). The resulting mice were then crossed to Flp transgenic mice to remove the neomycin resistance cassette and generate mice with a Glut3Flox allele, in which exon 6 is flanked by LoxP sites. Ly6gCre, also called Catchup mice (17), were obtained from M. Gunzer, University Duisburg-Essen, in a C57BL/6 background. G1, and Ly6gCre mice were bred with KPFrt mice to obtain the following genotypes: KrasFrt-STOP-Frt-G12D/WT; Tp53Frt/Frt; Ly6gCre/WT; Glut1Flox/Flox, KrasFrt-STOP-Frt-G12D/WT; Tp53Frt/Frt; Ly6gWT/WT; Glut1Flox/Flox, or KrasFrt-STOP-Frt-G12D/WT; Tp53Frt/Frt; Ly6gCre/WT; Glut1WT/WT, the latter two being used as control (no Cre, or Glut1WT/WT). Tumors were initiated upon intratracheal instillation of 107 plaque-forming units (PFU) per mouse of a commercially available adenoviral CMV-Flp vector (Ad5CMVFlpo, University of Iowa Viral Vector Core Facility) to activate oncogenic KrasG12D and delete Tp53. To follow tumor development, mice were anesthetized using isoflurane (Piramal, 56.761.002) and maintained under anesthesia during the scanning procedure. Lungs were imaged with an X-Ray microtomography (Quantum FX; PerkinElmer) at a 50 μm voxel size, with retrospective respiratory gating. Individual tumor volumes were measured and calculated using Analyze 12.0 (PerkinElmer) or Osirix MD (Pixmeo, RRID:SCR_013618). For the BrdU assay, 2 mg of freshly prepared BrdU (Merck, 10280879001) was injected intraperitoneally into 100 μL of PBS 1×. Immunotherapy treatment was performed with an intraperitoneal injection of InVivoMAb anti-mouse PD-1 (clone 29F.1A12, Bio X Cell, BE0273, RRID:AB_2687796) at a dose of 200 μg/mouse twice a week for 2 weeks. Anti G-CSF treatment was performed with an intraperitoneal injection of 10 μg anti-mouse G-CSF antibody (clone 67604, R&D Systems, MAB414, RRID:AB_2085954) every day for 14 days. G-CSF treatment was performed with daily intraperitoneal injection of 10 μg of recombinant mouse G-CSF (PeproTech, 250-05). Tumor-bearing mice were subjected to radiotherapy with a 20 mm2 collimator (13 mA, 3 mm Cu filter, 225 keV) 11.6 Gy with a single dose delivered in 256 seconds as described previously (9). Experimental melanoma lung metastases were obtained by injecting 3 × 105 B16-F1 cells (RRID:CVCL_0158) into the tail vein of female syngeneic recipient mice, either control (Ly6gWT/WT; Glut1Flox/Flox or Ly6gCre/WT; Glut1WT/WT) or Glut1KO (Ly6gCre/WT; Glut1Flox/Flox). All lungs were collected, fixed, and paraffin-embeded 21 days after the injection. Hematoxylin and eosin (H&E) staining was performed on sections to detect the experimental metastases, which were counted blindly.
All mouse experiments were performed with the permission of the Veterinary Authority of Canton de Vaud, Switzerland (license number: VD2391).
Tumor and tissue-derived supernatant production
KPFrt tumors and healthy tissues (lung and spleen) were isolated and dissociated as described previously (10). The cell suspension was then cultured in complete DMEM (Thermo Fisher Scientific, 41965062) supplemented with 10% FBS (Thermo Fisher Scientific, 10270106) for 24 hours at a concentration of 107 cells/mL. The day after, supernatant was collected, filtered (0.22 μm), aliquoted after centrifugation (5 minutes, 2,000 rpm), and stored at −80°C or used immediately. For CD45+ and CD45− fraction-derived supernatant, CD45+ cells were sorted using CD45 MicroBeads (clone 30F11.1; Miltenyi Biotec, 130-052-301, RRID:AB_2877061), using the manufacturer’s instructions. Both positive and negative fractions were cultured for 24 hours at a concentration of 107 cells/mL.
Mouse neutrophil isolation
Tumor-associated neutrophils (TAN), healthy lung neutrophils (HLN), and bone marrow neutrophils were isolated using anti-Ly6G MicroBead Kit (Miltenyi Biotec, 130-092-332) and anti-Ly6G MicroBeads UltraPure (clone REA526, Miltenyi Biotec, 130-120-337) according to the manufacturer’s instructions. Where indicated, bone marrow neutrophils were stimulated with TNF (PeproTech, 315-01A, 10 ng/mL, 4 hours) before RNA isolation.
Neutrophil in vitro survival assay
A total of 105 freshly-isolated neutrophils were cultured for 20 or 48 hours in 200 μL of complete DMEM (Thermo Fisher Scientific, 41965062, 10270106) or complete DMEM supplemented with 16% of tumor-derived supernatant. For size exclusion separation of proteins from the tumor-derived supernatant, the Centricon Plus-70 Centrifugal Filter (Merck, UFC700308) was used according to the manufacturer’s instructions.
The SV2 cell line was generated in our laboratory in 2013 from a lung tumor of a KP mouse; it was passaged >20 times before use. Around the time of the experiments, SV2 cells were not tested for mycoplasma. The B16-F1 cells were obtained in 2020 from D. Constam (ISREC, EPFL) and were mycoplasma negative.
Neutrophil-tumor cell coculture to measure tumor cell growth
100, 200, or 400 KP-derived cells (SV2) were cultured alone or in presence of 104 control or Glut1KO TANs. After 10 days in culture, cells were fixed for 10 minutes at room temperature in 3.6% formaldehyde (VWR Chemicals, 20909.290) and colored with crystal violet (Merck, C3886) for 5 minutes at room temperature. The percentage of covered area was then measured and quantified using Fiji (RRID:SCR_002285).
CD45 negative cell enrichment
Individual tumors were isolated and dissected and CD45+ cells were depleted using CD45 MicroBeads (clone 30F11.1, Miltenyi Biotec, 130-052-301, RRID:AB_2877061), using the manufacturer’s instructions.
RNA isolation and real-time PCR
Total RNA was extracted using TRIzol Reagent (Invitrogen, 15596018) according to the manufacturer’s instructions. cDNAs were synthesized from 1 μg total RNA using the High-Capacity cDNA Reverse Transcription Kit (Thermo Fisher Scientific, 4368814). Real-time PCR was performed using 5 ng of cDNA. The gene quantification was performed using the following commercially available probes from Thermo Fisher Scientific: Rpl30 (Mm01611464_ g1, used for normalization), Glut1 (Slc2a1) (Mm00441480_m1), Glut3 (Slc2a3) (Mm00441483_m1), Arg1 (Mm00475988_m1), Hk2 (Mm00443385_ m1), Siglecf (Siglec5) (Mm00523987_m1), Ikbke (Mm00444862_m1), Apoe (Mm01307193_g1), Pfkp (Mm00444792_m1), Cxcl3 (Mm01701838_ m1), Grx1 (Mm01352826_ g1), Il23a (Mm00518984_m1), Padi4 (Mm01341658_m1), and G-CSF (Csf3) (Mm00438334_m1). Ct values were analyzed according to the comparative Ct method.
RNA from the Ly6G-positive or CD45− fractions was extracted using TRIzol. Multiplexed libraries for mRNA-seq were prepared following SMARTer Stranded Total RNA-seq Pico v2 prep, starting from 1 ng total RNA from the Ly6G-positive fractions; and following TruSeq stranded mRNA LT prep, starting from 375 ng of RNA from CD45− fractions. Sequencing was subsequently performed on a NextSeq 500 instrument (Illumina) on a high-output flow cell, yielding single-end reads of 75 nucleotides. Adapter sequences and low-quality ends were removed with cutadapt (v1.14) as needed. Reads were aligned to mouse genome build mm10 using HISAT2 aligner (v2.10). Genes with low expression were filtered out (average transcripts per kilobase million <6 or average raw counts <50). Counts were normalized for library size using TMM method from EdgeR (v3.24.3) and voom from limma. Differential expression was computed with limma between TANs and HLNs (significance cutoff LFC >1 or < −1 and adjusted P value <0.05), Glut1KO and control TANs, Glut1KO and control non-immune fraction (significance cutoff LFC >1 or < −1 and P value <0.05).
We investigated over-representation of pathways as defined in the Hallmark collection of MSigDB (v6.0) among the top 200 genes upregulated in TANs versus HLNs or downregulated in Glut1KO versus control TANs.
Western blots were performed on protein extracts from freshly isolated neutrophils. Cells were lysed in RIPA buffer [20 mmol/L Tris pH8, 50 mmol/L NaCl, 0.5% sodium deoxycholate, 0.1% SDS, 1 mmol/L Na3VO4, protease inhibitor cocktail (complete, Roche)] at 4°C for 30 minutes. BCA Protein Assay Kit (Merck, 71285-3) was used for protein quantification. Anti-GLUT1 (Millipore, 07-1401, RRID:AB_1587074) and anti-Histone H3 antibodies (Abcam, ab1791, RRID:AB_302613) were used.
Flow cytometry analysis of tumors was performed as described previously (10). Antibodies and fluorochromes used were the following: anti-CD45-BUV661 (clone 30-F11, BD Biosciences, 565079, RRID:AB_2739057), anti-CD11b-BV711 (clone M1/70, BioLegend, 101242, RRID:AB_2563310), anti-Ly6G-PE (clone 1A8, BioLegend, 127607, RRID:AB_1186104), anti-Siglec-F-PE-Vio 615 (clone REA798, Miltenyi Biotec, Catalog No. 130-112-172, RRID:AB_2653444), anti-BrdU-APC (clone Bu20a, BioLegend, 339808, RRID:AB_10895898), anti-PD-L1-BV-785 (clone 10F.9G2, BioLegend, 124331, RRID:AB_2629659). Samples were acquired using the LSRII SORP (Becton Dickinson), a 5-laser and 18-detector analyzer at the EPFL Flow Cytometry Core Facility. CountBright Absolute Counting Beads (Thermo Fisher Scientific, C36950) were used for Trucount experiments using the manufacturer’s instructions. Data analyses were performed using FlowJo (FlowJo LLC, RRID:SCR_008520).
IHC and immunofluorescence
Preparation of mouse lung sections was explained before (10). Antibodies used for IHC from mouse tissue sections are anti-GLUT1 (Abcam, ab115730, RRID:AB_10903230, 1:650), anti-pHH3 (Cell Signaling Technology, 9701S, RRID:AB_331535, 1:100), anti-pERK (Cell Signaling Technology, 4370S, RRID:AB_2315112, 1:400), and anti-S100A9 (Novus Biologicals, NB110–89726, RRID:AB_1217846, 1:5,000). For human sample multiplexing, the 4plex immunofluorescence was performed using the fully automated Ventana Discovery ULTRA (Roche Diagnostics). All steps were performed on the instrument with Ventana solutions following their recommendations. Briefly, dewaxed and rehydrated paraffin sections were pretreated with heat using standard conditions for 40 minutes in CC1 solution. Primary antibodies were rabbit anti-GLUT1 (Millipore, 07-1401, RRID:AB_1587074, 1:5,000), rabbit anti-CD4 (clone SP35, Roche Diagnostics, 05552737001), rabbit anti-CD8 (clone SP57, Roche Diagnostics, 05937248001), rabbit anti-S100A9 (Novus Biologicals, NB110-89726, RRID:AB_1217846, 1:20,000). They were applied sequentially and incubated during 1 hour (GLUT1 and S100A9) or 20 minutes (CD4 and CD8) at 37°C (the data obtained with CD4 and CD8 staining were not used in this study). After incubation with an anti-rabbit Immpress HRP (ready to use, Vector Laboratories, MP-7401, RRID:AB_2336529), they were sequentially revealed with the Rhodamine-6G (Roche Diagnostics, 07988168001), DCC (Roche Diagnostics, 07988192001), Red 610 (Roche Diagnostics, 07988176001), and Cyanine 5 (Roche Diagnostics, 07551215001) TSA kits. Heat denaturation was applied after each TSA revelation. Slides were then acquired with an Olympus slide scanner (Olympus, VS120-L100) and analyzed using QuPath (RRID:SCR_018257). Permission to use the human material was granted by the local ethics committee (KEK: 200/2014).
Extracellular acidification rate (ECAR) measurements were performed using XF96 Extracellular Flux analyser (Seahorse Bioscience). Briefly, 100,000 freshly isolated neutrophils from healthy lungs or KP tumors were plated into a XF96 V3 polystyrene cell culture plate (Seahorse Bioscience). Cells were incubated for 3 hours in a humidified 37°C incubator with 5% CO2 in RPMI with no glucose (Thermo Fisher Scientific, 11879020). All experiments were performed at 37°C. Each measurement cycle consisted of a mixing time of 3 minutes and a data acquisition period of 2 minutes. For glycolysis assay, 10 mmol/L glucose (VWR, 1.08337.0250) was added during the seahorse experiment followed by 50 mmol/L 2-deoxyglucose (2-DG; Merck, D8375-5G). At the end of the experiment, proteins were extracted using RIPA buffer [20 mmol/L Tris pH8, 50 mmol/L NaCl, 0.5% sodium deoxycholate, 0.1% SDS, 1 mmol/L Na3VO4, protease inhibitor cocktail (complete, Roche)] to determine protein concentration in each well. ECAR was calculated as normalized against protein concentration.
2-NBDG glucose uptake experiments were performed on freshly isolated neutrophils using a Glucose-Uptake Cell-Based Assay Kit (Cayman Chemical, CAY-600470–1) according to the manufacturer’s instructions. Data were normalized against protein concentration.
ATP production assay
Production of ATP in freshly isolated neutrophils was measured using the CellTiter-Glo Luminescent Cell Viability Assay (Promega, G7570) according to the manufacturer’s instructions. Data were normalized against protein concentration.
Mice were age-matched and randomized according to tumor progression and gender. For metabolic and survival assays, μCT follow-up, flow cytometry, IHC, and real-time PCR, the samples were processed and analyzed blindly. All results are represented as mean ± SD unless stated otherwise. Each point represents an independent tumor (for in vivo experiments) or an independent well (for ex vivo experiments). Comparisons between groups were made as indicated in the figure legends. Statistical significance is indicated as *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001; or ns (not significant), using two-sided Mann–Whitney tests unless specified otherwise. Statistical analysis was performed using GraphPad Prism (version 6, RRID:SCR_002798).
The RNA sequencing data have been deposited to the GEO database (https://www.ncbi.nlm.nih.gov/geo/) and assigned the identifier: GSE140159.
Glucose transporter GLUT1 mediates neutrophil survival
To address how the tumor microenvironment may alter neutrophils, we first compared the fate of freshly-isolated bone marrow-derived neutrophils, cultured either in complete medium or in complete medium supplemented with a KP lung tumor-derived supernatant (Supplementary Fig. S1A). After 20 or 48 hours in culture, the medium supplemented with tumor-derived supernatant increased neutrophil survival significantly (Supplementary Figs. S1B and S1C). Size exclusion separation of proteins revealed that only the protein fraction of >3 kDa increases neutrophil survival (Supplementary Fig. S1D), excluding a prosurvival effect mediated by pH or small metabolites. We next generated tumor supernatants from CD45+ (immune) or CD45− (non-immune) cells. The CD45−-derived supernatant was sufficient to increase neutrophil survival, whereas the CD45+ supernatant had no effect (Supplementary Fig. S1E). Thus, this experimental system, which enables to extend neutrophil lifespan ex vivo, could be used to identify mechanisms altering their survival in tumors.
To determine if neutrophil survival is enabled by an increased metabolic activity, we next monitored the expression of two high affinity glucose transporters, Glut1 and Glut3, in response to tumor-derived supernatant. Surprisingly, the two transporters were regulated in an opposite manner, with Glut1 being upregulated and Glut3 downregulated upon incubation with tumor-derived supernatant (Fig. 1A).
To directly test the importance of GLUT1 for neutrophil survival, we crossed Ly6GCre mice, where the Cre recombinase gene is knocked into the Ly6g allele (17), to Glut1Flox/Flox mice (16), generating mice with neutrophil-specific Glut1 deletion (Fig. 1B). In the blood of healthy mice, both control and Glut1KO neutrophils represent between 8% and 10% of the immune cell population (Supplementary Fig. S2A). We isolated Glut1KO and control neutrophils, and monitored a partial-to-complete loss of survival benefit upon tumor-derived supernatant incubation in the Glut1KO cells (Fig. 1B and C; Supplementary Fig. S2B). Altogether, our ex vivo data indicate that soluble factors produced in lung tumors increase neutrophil survival through GLUT1 upregulation, a response that may favor an extended neutrophil lifespan within the tumor microenvironment.
GLUT1 and glucose metabolism are enhanced in TANs
To determine if GLUT1 is expressed in TANs from human LUAD, we used immunofluoresecence and stained tumor sections with anti-S100A9 to mark neutrophils, and with anti-GLUT1. We identified both GLUT1 positive and GLUT1 negative human TANs, which suggests a metabolic heterogeneity for intratumoral neutrophils (Fig. 2A).
To better understand the metabolic alterations of TANs in vivo, we isolated neutrophils from age-matched healthy lung (HLNs) and from KP tumors (TANs; Fig. 2B). RNA sequencing analyses identified 1470 upregulated genes and 487 downregulated genes in TANs compared to HLNs (Supplementary Table S2). Pathway analysis revealed glycolysis as the most significantly upregulated pathway in TANs (Supplementary Figs. S3A and S3B; Supplementary Table S3). By real-time PCR analysis, we validated the TAN signature with multiple genes from the lists of up- and downregulated genes (Supplementary Fig. S3C). The mRNA expression of each of Glut1 and Glut3 was stronger in neutrophils compared with other immune cell populations isolated from healthy lung or spleen (Fig. 2C) and, similar to the tumor supernatant experiments (see Fig. 1A), Glut1 was higher and Glut3 lower in TANs compared with HLNs (Fig. 2D and E). The strong GLUT1 expression in TANs was particularly evident at protein level (Fig. 2F). Furthermore, the expression of the glycolytic gene, hexokinase 2 (Hk2), was significantly more elevated in TANs than in HLNs (Supplementary Fig. S3C). Congruent with the upregulated glycolysis gene signature, including stronger Hk2 and Glut1 expression, we demonstrated that TANs undergo a metabolic switch toward increased glycolysis, because they exhibit an elevated extracellular acidification rate in response to glucose (Fig. 2G), absorb glucose more efficiently (Fig. 2H) and produce more ATP (Fig. 2I) compared with HLNs. Thus, when compared to HLNs, lung TANs are characterized by a shift toward increased glucose usage and metabolism.
An accelerated turnover defines GLUT1-deficient TANs
To directly interrogate the importance of GLUT1 in TANs, we crossed a second-generation KrasFrt-STOP-Frt-G12D/WT; Tp53Frt/Frt (referred to as KPFrt thereafter) tumor mouse model (15) to Ly6gCre knockin and Glut1Flox/Flox (16, 17) conditional knockout mice (Fig. 3A). In this experimental configuration, intratracheal instillation of adenoviral-Flp activates oncogenic Kras (G12D) and deletes Tp53, resulting in tumor development in a context where neutrophils and TANs lack Glut1.
First, we used immunocytochemistry and Western blot analysis to confirm the loss of GLUT1 in TANs at protein level (Fig. 3B). Contrary to our expectations, there was a higher prevalence of Glut1KO neutrophils in the tumor mass compared with control neutrophils (Fig. 3C; Supplementary Fig. S3D); however, their absolute numbers based on the quantification of S100A9 staining by immunofluorescence were not increased (Fig. 3D). Next, we monitored the molecular changes consequent to Glut1 loss in TANs using RNA sequencing (Supplementary Table S4). The downregulation of the “TNFα signaling via NF-κB” gene signature in Glut1KO TANs (Fig. 3E; Supplementary Table S5), which was higher in wild-type TANs compared with HLNs (see Supplementary Fig. S3B, left), might indicate that TNF and NF-κB signaling in TANs is linked to GLUT1-dependent survival (18, 19). This possibility was illustrated by an induction of Glut1 in response to TNF stimulation of bone marrow-derived neutrophils (Supplementary Fig. S3B, right).
A SiglecFhigh neutrophil subset was recently reported to have an increased lifespan in lung adenocarcinoma (20) and in myocardial infarction (21). In Glut1KO TANs, the SiglecFhigh subset was very significantly decreased (Fig. 3F). To further explore this subset, we analyzed data from different neutrophil subpopulations recently identified by single cell RNA sequencing in another KP-derived model, where six subsets of neutrophils were identified, in blood (N1 and N2) and in tumors (N3–N6; ref. 22). We observed that the SiglecFhigh cells were predominant in the N4, tumor-specific subpopulation. In addition, N4 was defined by high Glut1, Hk2, and Pd-l1, and by low Glut3 expression (Supplementary Fig. S4A), highlighting the relevance of GLUT1 for this TAN subset. From autochtonous KP tumors, we demonstrated that Glut3 but not Glut1 expression was significantly reduced in the SiglecFhigh compared with the SiglecFlow neutrophil subpopulation, leaving GLUT1 as the predominant glucose transporter of SiglecFhigh TANs (Supplementary Fig. S4B).
To investigate the mechanisms by which neutrophil turnover might be enhanced upon Glut1 deletion, we performed an RNA sequencing of the non-immune fraction (CD45−) of the tumor mass. Among the highest upregulated genes in tumors having Glut1KO neutrophils, we identified several genes involved in neutrophil recruitment (Fig. 3G; Supplementary Table S6). In addition, although there is no significant difference in neutrophil percentage between the genotypes in healthy mice (see Supplementary Fig. S2A), we measured a higher proportion of blood-circulating neutrophils in control tumor-bearing compared to healthy mice, which was further increased in KPFrt; Ly6gCre/WT; Glut1Flox/Flox tumor-bearing mice (Fig. 3H). These results suggest an elevated neutrophil turnover during cancer growth with Glut1KO TANs. Neutrophil turnover was measured from different anatomical sites using an in vivo BrdU tracking experiment, in healthy and tumor-bearing mice. Specifically, bone marrow, blood, spleen, lung, and tumors were collected 2.5 and 6.5 days after a single BrdU injection (Fig. 4A). In all healthy tissues, the BrdU kinetics was similar, with a peak (colonization phase) at 2.5 days reaching nearly 50% BrdU-positive neutrophils (Fig. 4A–C). After 6.5 days, the majority of neutrophils were BrdU negative, showing that the entire pool had been renewed (extinction phase; Fig. 4A–C). As exception, in lungs a small fraction (10%) of neutrophils was still BrdU positive 6.5 days after injection. In contrast, in lung tumors the kinetics was altered, with only 10% of BrdU-positive neutrophils after 2.5 days, rising up to 30% after 6.5 days, thus still being in the colonization phase. These data demonstrate a reduced neutrophil turnover specifically in the tumor mass, leading to an aberrantly large proportion of old (6.5 days) neutrophils (Fig. 4A–C). In tumors, at 2.5 days all the young BrdU-positive neutrophils (<2.5 days-old) were SiglecFlow, whereas, at 6.5 days, the old BrdU-positive TANs were mainly SiglecFhigh (Fig. 4B), consistent with a time-dependent acquisition of this membrane protein in lung tumors.
When comparing tumors with Glut1KO neutrophils to controls, we identified that Glut1KO TANs have an accelerated turnover with more BrdU-positive cells at 2.5 days post-BrdU injection and a reduced old neutrophil fraction 6.5 days after (Fig. 4D). Thus, in tumors with Glut1KO neutrophils, reduced neutrophil survival and increased neutrophil recruitment together contribute to a higher neutrophil turnover during cancer progression. These results also position the old, SiglecFhigh and Glut1high TAN subset as an inhibitor of young neutrophil recruitment, possibly acting by a communication with the tumor epithelial cells.
Glut1KO but not Glut3KO neutrophils negatively impact tumor growth
Because the pro- and antitumor properties of neutrophils require a better understanding (23), we decided first to explore the consequences of Glut1KO TANs on tumor progression (Fig. 5A). Tumors with Glut1KO TANs had a reduced cell proliferation as measured by phospho-Histone H3 (pHH3) and a reduced phospho-ERK1/2 (pERK) staining, a marker of tumor progression (Fig. 5B and C). Moreover, tumor growth rates monitored by longitudinal micro-computed tomography (μCT) were significantly diminished in mice with Glut1KO TANs (Fig. 5D). Next, to determine if TANs affect lung tumor cells in vitro, we measured their behavior using a KP-derived cell line (SV2) cultured alone, with control TANs or with Glut1KO TANs. A stronger negative impact on tumor cell growth was observed in Glut1KO TANs, as revealed by a reduced spreading of cocultured SV2 cells (Fig. 5E). These suggest that Glut1KO TANs have a reduced tumor-supportive capacity or are endowed with antitumor properties.
To test if the reduced tumor-supportive function of Glut1KO TANs can be extended to other models, we injected B16-F1 melanoma cells via the tail vein of control or Glut1KO syngeneic recipient mice, and counted the number of lung lesions 3 weeks later. In this experimental metastasis model, there were fewer lesions in Glut1KO conditions (Fig. 5F). Furthermore, tumor cell proliferation was reduced in these tumors compared with control tumors, as indicated by a reduced proportion of pHH3 positive cells (Fig. 5G). Thus, TAN-mediated tumor support relies on GLUT1 in different cancer types.
Intrigued by the opposite regulation of Glut1 and Glut3 in response to tumor-derived supernatant, in HLNs versus TANs, and by their different expression patterns in the neutrophil subsets (see Figs. 1A, 2D and E; Supplementary Fig. S4A), we decided to interrogate directly the consequences of Glut3 gene deletion in neutrophils (Supplementary Figs. S5A and S5B). In contrast to our data obtained with Glut1KO, we monitored a reduced proportion of Glut3KO neutrophils in the blood of healthy mice (Supplementary Fig. S5C), no tumor growth delay in mice with Glut3KO neutrophils compared with controls (Supplementary Figs. S5D–S5F), and no change in the proportion of the SiglecFhigh TAN subset compared with controls (Supplementary Fig. S5G). Thus, GLUT1 but not GLUT3 deficiency in neutrophils diminishes tumor growth.
To understand the mechanisms by which the loss of GLUT1 in TANs reduces tumor progression, we hypothesized that the turnover of neutrophils may be critical to regulate the neutrophil subset balance. Specifically, considering that G-CSF affects neutrophil mobilization and survival (24), we decided to explore its role in this balance using loss- and gain-of-function approaches. Daily injections of G-CSF neutralizing antibodies in tumor-bearing mice significantly reduced the number of TANs (Fig. 6A and B). This decrease could be explained by a reduced mobilization or by a recruitment defect at the tumor site, as illustrated by the very significant loss of young BrdU+ neutrophils (2.5 days) without affecting the acquisition of SiglecF (Fig. 6C–E). This decreased number of the entire TAN pool led to a reduced tumor growth, highlighting the protumoral properties of these innate immune cells (Fig. 6F). Accordingly, G-CSF treatment reduced the proportion of SiglecFhigh TANs (Supplementary Figs. S6A–S6C), but only in Glut1KO conditions and increased the abundance of young (BrdU+ at 2.5 days) neutrophils in tumors with control and Glut1KO neutrophils (Supplementary Figs. S6D and S6E). Although G-CSF decreased the growth rate of control tumors, its impact was more pronounced in Glut1KO tumors (Supplementary Fig. S6F). Thus, the antitumor properties of young neutrophils seem to be amplified upon GLUT1 loss. Altogether, our data suggest the existence of a functional link between neutrophil turnover and tumor growth (Fig. 6G).
Glut1KO neutrophils sensitize lung tumors to radiotherapy
Because neutrophils have been previously linked to immuno- and radiotherapy responses (10, 25), we next wondered if GLUT1 deficiency in these cells could potentiate such treatments. Despite their reduced expression of PD-L1, Glut1KO neutrophils failed to sensitize KP tumors to anti-PD1 (Supplementary Figs. S7A and S7B). In contrast, the antitumor response to radiotherapy was markedly enhanced, with multiple (62.5%) regressing tumors monitored 2 weeks after irradiation, and a long-term (at least 28 days) loss of tumor growth specific to Glut1KO TAN-irradiated mice (Fig. 7A). Thus, GLUT1 deletion in neutrophils increases KP lung tumor sensitivity to radiotherapy, leading to a durable growth impairment.
In this study, we demonstrated that tumor-derived secreted factors increase neutrophil survival, which is enabled by glucose transporter GLUT1. Using a mouse model of lung adenocarcinoma, we identified a reduced neutrophil turnover specifically in the tumor compartment. We found that an important proportion of TANs can survive at least 6.5 days, providing an abnormally high level of old neutrophils when compared with other compartments including the healthy lung.
Within lung tumors, the aberrantly long neutrophil lifespan provides sufficient time to acquire tumor-associated markers such as SiglecF and PD-L1. Whether the acquisition of tumor-supportive properties is passive and linked to the reduced TAN turnover, or active and triggered by specific factors remains to be elucidated. Moreover, in vivo G-CSF treatment or neutralization failed to demonstrate a direct link with the acquisition of SiglecF, whereas its inhibition strongly impacted neutrophil infiltration in the tumor. On the contrary, G-CSF treatment led to a massive recruitment of young neutrophils, accompanied by a reduced tumor development. The impact of this cytokine on young neutrophil recruitment and tumor growth was exacerbated when neutrophils did not express GLUT1, reducing the proportion and number of old, SiglecFhigh neutrophils. These results suggest the existence of functionally diverse and competing neutrophils in tumors. A facet of this putative competition was observed when the old SiglecFhigh subpopulation decreased as a consequence of GLUT1 loss. When TANs lack GLUT1, neutrophil chemoattractants are induced in tumors, which may contribute to an enhanced recruitment of young neutrophils, a response that is accompanied by a reduction of tumor growth. Such a systemic response could resemble one described for TRAIL receptor (TRAIL-R)-mediated apoptosis in myeloid-derived suppressor cells (26); alternatively an increased homing mediated by local chemoattractants might occur upon GLUT1 deficiency. The precise mechanisms and the cells responsible in this increased recruitment remain to be elucidated.
Metabolic alterations in the tumor immune compartment have been described in multiple cell types (27). Although here we focused on glucose transporters in TANs, another study using a 4T1 mammary tumor cell line transplantation model reported a switch toward oxidative metabolism (28) in cKit positive, immature TANs. In addition, bone marrow neutrophils from mice with early-stage cancer were characterized by enhanced spontaneous migration compared with neutrophils from tumor-free mice (29). These cells had a higher rate of glycolysis and ATP production compared with control neutrophils, suggesting that some neutrophils are prone to rewire their glucose metabolism directly at the bone marrow stage. Undoubtedly, more work will be needed to characterize and compare how neutrophil subsets vary metabolically and functionally in different tumor types and during tumor evolution.
The accelerated turnover of neutrophils in Glut1KO or under G-CSF treatment was sufficient to reduce tumor growth significantly. However, the alterations of tumor progression were visible only after a few weeks of monitoring, and we did not observe tumor regression. In contrast, the loss of GLUT1 augments radiotherapy efficacy, leading to a tumor regression never observed previously with different radiotherapy protocols (this study and refs. 25, 30). Upon radiotherapy, proliferating cancer cells accumulate double-strand breaks and die. The antitumor role of neutrophils seems to rely on direct cancer cell killing (31), which might be stimulated after radiotherapy treatment where cells are stressed and exposed to a strong local inflammation. The link between inflammation, antitumor neutrophils and radiotherapy treatment needs to be better understood to envisage possible future neutrophil-based strategies against cancer.
In conclusion, our work highlights the importance of enhanced glucose uptake and metabolism for TANs, enabling tumor support. Although more experiments will be needed to test this directly, our data favor the hypothesis that the presence of old and possibly protumor neutrophils competes with young antitumor neutrophils. We suggest that pro- or antitumor properties of neutrophils are a function of their turnover, glucose transporter expression, and differentiation status (Fig. 7B), a model that may reconcile earlier studies describing opposite roles for neutrophils in cancer. In such a scenario, therapeutic agents reducing neutrophil survival or increasing young neutrophil recruitment to the tumor may counteract tumor growth when used alone or in combination with conventional anticancer treatment modalities such as radiotherapy. However, a limitation of our work is that the data linking TAN survival to tumor support are mainly indirect, and firm proof for this link should be obtained from future investigations.
Finally, our study positions glucose metabolism and particularly GLUT1/3-mediated glucose uptake as a crucial node for TAN turnover that may control the equilibrium between pro- and antitumor neutrophils in lung cancer.
S. Peters reports personal fees from Abbvie, Amgen, Astra Zeneca, Bayer, Beigene, Biocartis, Boehringer Ingelheim, BMS, Clovis, Daiichi Sankyo, Debiopharm, Eli Lilly, Roche/Genentech, Foundation Medicine, Illumina, Incyte, Janssen, Medscape, MSD, Merck Seono, Merrimack, Novartis, Pharmamar, Phosphopatin Therapeutics, Pfizer, Regeneron, Sanofi, Seattle Genetics, and Takeda outside the submitted work. D.G. Kirsch reports other support from Lumicell, Xrad Therapeutics, Amgen, and Calithera; grants from Merck, Bristol-Myers Squibb, Varian Medical Systems outside the submitted work. J.C. Rathmell reports personal fees from Sitryx, Caribou, Nirogy, and Pfizer; grants from Incyte, Kadmon, Calithera, and Tempest outside the submitted work. No disclosures were reported by the other authors.
P-B. Ancey: Conceptualization, formal analysis, investigation, methodology, writing–original draft. C. Contat: Formal analysis, investigation, writing–review and editing. G. Boivin: Investigation, methodology. S. Sabatino: Investigation. J. Pascual: Investigation. N. Zangger: Data curation, formal analysis. J.Y. Perentes: Funding acquisition. S. Peters: Funding acquisition. E.D. Abel: Resources. D.G. Kirsch: Resources. J.C. Rathmell: Resources. M-C. Vozenin: Formal analysis. E. Meylan: Conceptualization, formal analysis, supervision, funding acquisition, project administration, writing–review and editing.
P-B. Ancey, S. Sabatino, J. Pascual, J.Y. Perentes, S. Peters, and E. Meylan were supported by a grant from the Swiss Cancer Research Foundation (KFS-4555-08-2018). P-B. Ancey, C. Contat, G. Boivin, and E. Meylan were supported by a grant from the Swiss National Science Foundation (310030_179324). C. Contat, N. Zangger, and E. Meylan were supported by the Emma Muschamp Foundation. D.G. Kirsch was supported by R35CA197616 from the NCI in the United States. We thank the EPFL SV Flow Cytometry, Gene Expression Core and Histology Core Facilities for help with flow cytometry, RNA sequencing, tissue preparation, and multiplexing, respectively. We thank S. Berezowska, I. Zlobec, and the Translational Research Unit of the Institute of Pathology (University of Bern, Switzerland) for generously providing the human LUAD samples. We thank D. Constam for providing the B16-F1 cells and G. Diaceri for injecting them (both from EPFL, Switzerland). We thank M. Pittet (MGH, Boston, MA) for critically reading the manuscript.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.