ARTICLE

https://doi.org/10.1038/s41467-020-17644-0 OPEN Immunoprophylactic and immunotherapeutic control of hormone receptor-positive breast cancer

Aitziber Buqué 1,2, Norma Bloy1,2, Maria Perez-Lanzón2,3,4, Kristina Iribarren2,4, Juliette Humeau2,3,4, Jonathan G. Pol2,4, Sarah Levesque2,3,4, Laura Mondragon2,4, Takahiro Yamazaki 1, Ai Sato 1, Fernando Aranda 2,4, Sylvère Durand2,4, Alexandre Boissonnas5, Jitka Fucikova6,7, Laura Senovilla2, David Enot2,4, Michal Hensler6, Margerie Kremer2,4, Gautier Stoll 2,4, Yang Hu 8, Chiara Massa9, Silvia C. Formenti 1,10, Barbara Seliger9, Olivier Elemento8,10, Radek Spisek6,7, Fabrice André 11, ✉ Laurence Zitvogel3,11,12,13, Suzette Delaloge 14, Guido Kroemer 2,4,15,16,17,20 & 1,8,10,18,19,20✉ 1234567890():,; Lorenzo Galluzzi

Hormone receptor (HR)+ breast cancer (BC) causes most BC-related deaths, calling for improved therapeutic approaches. Despite expectations, immune checkpoint blockers (ICBs) are poorly active in patients with HR+ BC, in part reflecting the lack of preclinical models that recapitulate disease progression in immunocompetent hosts. We demonstrate that mam- mary tumors driven by medroxyprogesterone acetate (M) and 7,12-dimethylbenz[a] anthracene (D) recapitulate several key features of human luminal B HR+HER2− BC, including limited immune infiltration and poor sensitivity to ICBs. M/D-driven oncogenesis is accelerated by immune defects, demonstrating that M/D-driven tumors are under immu- nosurveillance. Safe nutritional measures including nicotinamide (NAM) supplementation efficiently delay M/D-driven oncogenesis by reactivating immunosurveillance. NAM also mediates immunotherapeutic effects against established M/D-driven and transplantable BC, largely reflecting increased type I interferon secretion by malignant cells and direct stimu- lation of immune effector cells. Our findings identify NAM as a potential strategy for the prevention and treatment of HR+ BC.

1 Department of Radiation Oncology, Weill Cornell Medical College, New York, NY, USA. 2 Equipe labellisée par la Ligue contre le cancer, Centre de Recherche des Cordeliers, INSERM U1138, Université de Paris, Sorbonne Université, Paris, France. 3 Faculté de Médecine, Université de Paris Sud, Paris- Saclay, Le Kremlin-Bicêtre, Paris, France. 4 Metabolomics and Cell Biology Platforms, Gustave Roussy Comprehensive Cancer Institute, Villejuif, France. 5 Sorbonne Université, Inserm, CNRS, Centre d’Immunologie et des Maladies Infectieuses CIMI, Paris, France. 6 Sotio, Prague, Czech Republic. 7 Department of Immunology, Charles University, 2nd Faculty of Medicine and University Hospital Motol, Prague, Czech Republic. 8 Caryl and Israel Englander Institute for Precision Medicine, New York, NY, USA. 9 Institute of Medical Immunology, Martin Luther University Halle-Wittenberg, Halle, Germany. 10 Sandra and Edward Meyer Cancer Center, New York, NY, USA. 11 Gustave Roussy Cancer Center, Villejuif, France. 12 INSERM U1015, Villejuif, France. 13 Center of Clinical Investigations in Biotherapies of Cancer (CICBT) 1428, Villejuif, France. 14 Department of Cancer Medicine, Gustave Roussy Cancer Center, Villejuif, France. 15 Pôle de Biologie, Hôpital Européen Georges Pompidou, AP-HP, Paris, France. 16 Suzhou Institute for Systems Medicine, Chinese Academy of Medical Sciences, Suzhou, China. 17 Department of Women’s and Children’s Health, Karolinska University Hospital, Stockholm, Sweden. 18 Department of Dermatology, Yale School of Medicine, New Haven, CT, USA. 19 Université de Paris, Paris, France. 20These authors jointly supervised: Guido Kroemer, Lorenzo ✉ Galluzzi. email: [email protected]; [email protected]

NATURE COMMUNICATIONS | (2020) 11:3819 | https://doi.org/10.1038/s41467-020-17644-0 | www.nature.com/naturecommunications 1 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-17644-0

ormone receptor (HR)+ breast cancer (BC) causes the with a slow-release pellet liberating MPA (day 0) and received 6 majority of BC-related deaths in the US and Europe1. oral gavages of the environmental pollutant DMBA (administered H + Most newly diagnosed patients with HR BC patients are 1, 2, 3, 5, 6, and 7 weeks after implantation of the MPA pellet), managed by initial surgery, followed by adjuvant endocrine they developed mammary tumors with a highly variable latency therapy ± radiotherapy. Although gene expression profiles are (day ~50–150) with a 50% tumor incidence around day 90–110 used to predict benefit from adjuvant anthracycline-, cyclopho- (Fig. 1a, b). Similar findings were recorded for BALB/c female sphamide- and taxane-based chemotherapy2, risk assessment mice (Supplementary Fig. 1a). M/D-driven mammary tumors remains imprecise, and treatment efficacy is variable, resulting in occurred at a slightly higher incidence at inguinoabdominal vs many women being treated to benefit a small number. The cervicothoracic sites (Supplementary Fig. 1b), initially as single recognition that the individual benefits of chemotherapy are lesions but later as multifocal disease (at least in a fraction of limited and that current therapies are associated with consider- cases, Supplementary Fig. 1c), and progressed rapidly if untreated able side effects3 has shifted BC research toward novel approa- (Fig. 1b). ches, including immunotherapy. M/D-driven tumors often displayed: (1) a relatively mono- Immune checkpoint blockers (ICBs) have been successfully tonous population of cells, poorly circumscribed, infiltrating the implemented in the management of multiple tumors4,5. Accu- surround soft and adipose tissues in a loose pattern, (2) cords and mulating evidence indicates that the immune system also plays a nodules of atypical epithelial cells, with some duct or gland major role in the control of mammary carcinogenesis and tumor formation, and (3) some degree of vascular and lymphatic progression6. Accordingly, ICBs targeting programmed cell death permeation (Fig. 1c). Thus, M/D-driven tumors histologically 1 (PDCD1; best known as PD-1) or CD274 (the main PD-1 resemble to human HR+ BCs. Transcriptomic studies based on ligand, best known as PD-L1) have recently been suggested to The Cancer Genome Atlas (TCGA) database for human BC as a constitute good therapeutic options for triple-negative BC reference revealed that M/D-driven tumors display a striking (TNBC) patients7. However, the clinical experience with ICBs in similarity to highly proliferative HR+HER2− (luminal B) human HR+ BC patients has been disappointing8. At least in part, this BC (Fig. 1d and Supplementary Fig. 1d). Immunofluorescence reflects the lack of adequate preclinical models that recapitulate microscopy confirmed that untreated M/D-driven tumors express the incidence, natural progression, and response to therapy of estrogen receptor 1 (ESR1, best known as ER) but not vimentin HR+ BC in immunologically competent hosts. Indeed, human (VIM, a marker of basal BC) (Supplementary Fig. 1e). Accord- xenografts are intrinsically inadequate for studying tumor- ingly, M/D-driven tumors emerged with delayed kinetics in mice targeting immunity in mice, and mouse HR+ BC cell lines are bearing a deletion in Esr1 causing defective transcriptional implanted after being immunoedited by their original host (and activity (ER-AF20)13 (Fig. 1e), while oncogenesis and tumor hence have already evaded immune recognition)9. Similarly, progression were accelerated in mice expressing an ER mutant transgenic BC models often bear rather few non-synonymous associated with increased nuclear accumulation (ERC451A)14 mutations, and hence are largely invisible to the adaptive immune (Fig. 1f). The ER antagonist tamoxifen delayed M/D-driven system9. oncogenesis and cancer-related death, at least during the early In an attempt to circumvent these issues, we focused on stages of the process (Fig. 1g). Comparing the transcriptome of endogenous BC driven in mice by slow-release medrox- normal mouse mammary glands, hyperplastic glands exposed to yprogesterone acetate (MPA, M) pellets combined with an oral M/D but not developing tumors, and M/D-driven tumors carcinogen, 7,12-dimethylbenz[a]anthracene (DMBA, D)10. MPA revealed major changes linked to malignant transformation, is a multipurpose synthetic progestin that has been associated including the upregulation of proliferation-related genes as well with increased incidence of BC when used as part of hormone as in the downregulation of transcripts linked to immune replacement therapy in post-menopausal women11. DMBA is a functions (Fig. 1h). In particular, transcripts associated with polycyclic aromatic hydrocarbon (PAH) found in cigarette germinal centers (Btla, Cd4, Cd19, Cd22, Dock10, H2-M2) and smoke, industrial pollution and grilled meat, and exposure to tertiary lymphoid organs (Ccl21a, Ccl21b, Cd3d, Cd3e, Cd3g, these risk factors correlates with increased BC incidence12. Cxcl13, Dpp4, Il7r, Sell, Tcrb) were markedly downregulated Here, we report the results of in-depth biological, immunolo- during oncogenesis (Fig. 1h). Consistently, established M/D- gical, and functional studies demonstrating that M/D-driven driven tumors exhibited limited infiltration by CD3+CD8+ mammary tumors recapitulate several biological and immuno- cytotoxic T lymphocytes (CTLs) and CD3+CD4+ T cells, as logical features of human luminal B (highly proliferative compared to normal mammary glands (Fig. 1i), and their CD4+ HR+HER2−) BC, hence constituting a privileged platform to compartment was enriched in immunosuppressive FOXP3+ investigate therapeutic strategies with translational potential. regulatory T (TREG) cells (Fig. 1i). Thus, M/D-driven tumors Harnessing this model, we demonstrate that nicotinamide established in immunocompetent mice recapitulate key aspects of — fi (NAM) a variant of vitamin B3 currently sold over the counter human luminal B BC, which is also scarcely in ltrated by as nutritional supplement—not only delays the development of immune cells6. endogenous mammary carcinogenesis, but also exerts immu- notherapeutic effects against M/D-driven tumors and other transplantable mouse models of luminal B BC. The oncopre- M/D-driven tumors are under immunosurveillance by NK cells. ventive activity of NAM overcomes the detrimental effect of other Based on transcriptomic and cytofluorometric data (Fig. 1h, i), we common risk factors for BC, including a high-fat diet (HFD), and postulated that M/D-driven tumors would evolve by escaping relies on the reinstatement of T cell-dependent immuno- immunosurveillance. Indeed, M/D-driven carcinogenesis was surveillance, which is otherwise not operational during onco- accelerated, and overall survival (OS) was shortened, in Rag2−/− genesis. These findings identify NAM supplementation as a Il2rg−/− mice (lacking T cells, B cells and NK cells) (Fig. 2a), as potential strategy for the prevention and treatment of HR+ BC. well as in mice lacking interferon gamma (Ifng, encoding a major effector of lymphoid immunosurveillance) (Fig. 2b). To char- acterize the effects of immune components on the emergence of Results HR+ tumors, we performed a large-scale experiment in which we M/D-driven tumors recapitulate key features of human HR+ compared tumor-free survival (TFS) and OS following M/D-dri- BC. When 6–9-week-old female C57BL/6 mice were implanted ven oncogenesis between untreated (or sham-treated) wild-type

2 NATURE COMMUNICATIONS | (2020) 11:3819 | https://doi.org/10.1038/s41467-020-17644-0 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-17644-0 ARTICLE

a b 100 C57BL/6 (n = 19) 100 C57BL/6 (n = 19)

75 MPA DMBA Detection 75 50 50 TTD (%) TFS (%) 25 25

Day 0 71421 35 42 49 0 0 50 100 150 200 0 10 20 30 Days since 1st gavage Days after detection c Human Mouse

deERWT (n = 12) ER-AF2° (n = 12)

0.6 Probalility of 100 HR 0.24 [0.09;0.64] 100 HR 0.29 [0.12;0.77] Mouse molecular subtype p = 0.0018 p = 0.0084 Human 75 75 0.4

0.2 0.6 X1_T1_M8A1T X2_T2_M8A2T X3_T4_M8A4T X4_T5_M8B1T X5_T6_M8B4T X6_T7_M8C1Tp + – 50 50 ER /HER2 Low Prolif Density OS (%)

0.2 TFS (%) + – ER /HER2 High Prolif 25 25 HER2+ 0.0 0 0 –4 –2 0 2 4 ER–/HER2– 40 80 120 160 60 80 100 120 140 160 Expression Days since 1st gavage Days since 1st gavage

f ERWT (n = 10) ERC451A (n = 12) g C57BL/6 (n = 10) + Tamoxifen (n = 10)

100 HR: 3.86 [1.45;10.32] 100 HR: 2.74 [1.10;6.78] 100 HR: 0.51 [0.20;1.28] 100 HR 0.43 [0.16;1.11] p<0.0001 p = 0.0054 p = 0.0941 p = 0.0329 75 75 75 75

50 50 50 50 OS (%) OS (%) TFS (%) TFS (%) 25 25 25 25

0 0 0 0 40 80 120160 200 50 100 150200 250 40 80 120160 200 60120 180 240 Days since 1st gavage Days since 1st gavage Days since 1st gavage Days since 1st gavage h GO TERM p value SYMBOL logFC adj. p value SYMBOL logFC adj. p value Regulation of immune system process 2.34 E–33 Ccl21a –3.827 0.000797 Tcrb-J –2.358 0.000268 Ms4a1 –3.783 0.000004 lkzf3 –2.347 0.000012 Positive regulation of biological process 1.52 E–23 Cxcl13 –3.759 0.006389 Prkcq –2.298 0.000769 Regulation of immune response 4.66 E–22 Cr2 –3.682 0.000000 Cd3g –2.278 0.000126 Activation of immune response 3.72 E–20 Bank1 –3.176 0.000005 Enpep –2.270 0.000004 Sell –3.029 0.000205 Tespa1 –2.255 0.000087 Leukocyte activation 4.70 E–20 Cd36 –2.949 0.000000 Cd22 –2.221 0.000041 Positive regulation of response to stimulus 1.21 E–19 Slamf6 –2.908 0.000007 Cd79b –2.190 0.000780 ltk –2.679 0.000035 Cd8b1 –2.189 0.000281 Positive regulation of metabolic process 1.83 E–18 Stap1 –2.666 0.000514 Ebf1 –2.149 0.000001 Biological adhesion 3.30 E–18 Cd4 –2.633 0.000575 Gimap8 –2.135 0.000046 Lymphocyte activation 4.97 E–18 Orm1 –2.631 0.042508 Cd3e –2.122 0.005849 Tusc5 –2.623 0.000134 Gm4951 –2.102 0.002606 Immune response- regulating surface receptor 2.12 E–17 Timd4 –2.622 0.000066 H2-M2 –2.081 0.000033 Cd96 –2.580 0.000117 Gimap6 –2.069 0.000001 Mitotic cell cycle progress 2.85 E–60 Fcer2a –2.559 0.000030 Dpp4 –2.052 0.000109 Mitotic nuclear division 2.88 E–57 Trbv13-2 –2.541 0.000539 Cd69 –1.977 0.000582 Nuclear chromosome segregation 7.13 E–32 Aoc3 –2.494 0.000021 Pyhin1 –1.962 0.032641 ll7r –2.488 0.018206 Btla –1.930 0.010653 1.88 E–26 Microtubule cytoskeleton organization Cd19 –2.468 0.000046 Cd55 –1.864 0.004031 Cellular component organization or biogenesis 5.03 E–25 Dpt –2.412 0.000000 Dock10 –1.862 0.000090 Normal mammary gland Hyperplastic mammary gland M/D-driven mammary tumor

i 0.5 100 n = 7 Mammary glands

0.4 ) 80 M/D-driven mammary tumors + 6 total cells) 6 0.3 60 p = 0.0084 n = 7 n = 7 n = 17 0.2 p = 0.0211 40 Cells x10 n = 17 p = 0.0126 n = 7 p = 0.0313 n = 17 Cells (% of CD4 0.1 n = 17 20

(Normalized to 10 0 0 CD3+ CD4+ CD8+ FOXP3+

(WT) BALB/c or C57BL/6 mice and: (1) mice bearing a range of receptors (i.e., IL1R, IFNAR1); or (3) WT mice depleted of CD4+ different inherited defects in innate and adaptive immune func- and CD8+ cells, or NK cells. This experiment confirmed that M/ tions (genotypes: Casp1−/−, Il21−/−, Il22−/−, Il21r−/−, Il1r1−/−, D-driven carcinogenesis is under the control of IFNG-producing Myd88−/−, Nlrp3−/−, Rag2−/−); (2) WT mice treated with agents cells and revealed that the absence of Il21 or Il22, the blockade of that neutralize specific cytokines (i.e., IFNG, IL17) or cytokine IFNAR1, and the depletion of NKG2D+ cells also precipitate

NATURE COMMUNICATIONS | (2020) 11:3819 | https://doi.org/10.1038/s41467-020-17644-0 | www.nature.com/naturecommunications 3 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-17644-0

Fig. 1 M/D-driven tumors recapitulate key features of human HR+ breast cancer. a Schedule of oncogenic challenges for the induction of M/D-driven tumors and representative images of M/D-driven tumors established in WT C57BL/6 mice. Scale bar = 1 cm. b Tumor-free survival (TFS) and time-to- death (TTD) in WT C57BL/6 mice subjected to M/D-driven oncogenesis. Number of mice is reported. c Representative histology of M/D-driven tumors as compared to human HR+ breast cancers. Scale bar = 50 µm. d Probeset distribution comparing the transcriptomic profile of 6 M/D-driven mammary tumors established in C57BL/6 mice with that of human breast cancers from the TCGA public database. Probability of molecular subtyping is reported for each mouse tumors. e, f TFS and overall survival (OS) of WT C57BL/6 mice and C57BL/6 bearing ER mutations that cause nuclear exclusion (e) or nuclear accumulation (f) subjected to M/D-driven oncogenesis. Number of mice, hazard ratio (HR) and p values (two-sided log-rank) are reported. g TFS and OS of WT C57BL/6 mice subjected to M/D-driven oncogenesis in control conditions or along with tamoxifen administration in the drinking water. Number of mice, HR and p values (two-sided log-rank) are reported. h. Non-supervised hierarchical clustering of the transcriptomic profile of M/D-driven tumors established in WT C57BL/6 mice (n = 6), mammary glands exposed to M/D but not developing tumors (n = 6) and M/D-naïve mammary glands (n = 6). Gene Ontology analysis, fold change (FC) and adjusted, two-sided p values are reported. Red, upregulation. Yellow, downregulation. i Relative amount of CD3+, CD8+, CD4+ and CD25+FOXP3+ cells infiltrating M/D-driven tumors in C57BL/6 mice vs syngeneic M/D-naïve mammary glands. Results are means ± SEM plus individual data points. Number of mice and p values (unpaired, two-sided Student’s t, as compared to M/D-naïve mammary glands) are reported.

a WT (TFS, n = 11; OS, n = 11; TTD, n = 11) b Control (TFS, n = 9; OS, n = 9; TTD, n = 9) –/– –/– Rag2 Il2rg (TFS, n = 23; OS, n = 23; TTD, n = 21) Ifng–/– (TFS, n = 9; OS, n = 10; TTD, n = 9) 100 100 100 100 100 100 HR: 3.32 [1.43; 7.68] HR: 2.94 [1.29; 6.71] HR: 1.00 [0.48; 2.07] HR: 5.28 [1.87; 14.92] HR: 3.68 [1.38; 9.82] HR: 1.14 [0.45; 2.87] p = 0.0051 p = 0.0104 p = 0.9978 p = 0.0017 p = 0.0094 p = 0.7362 75 75 75 75 75 75

50 50 50 50 50 50 OS (%) OS (%) TTD (%) TFS (%) TTD (%) 25 25 25 TFS (%) 25 25 25

0 0 0 0 0 0 40 80 120 160 200 50 100 150 200 250 0 20 40 60 60 80100 120 140 160 60 80100 120 140 160 0102030 Days since 1st gavage Days since 1st gavage Days after detection Days since 1st gavage Days since 1st gavage Days after detection

cdANOVA p value; HR [95% CI] p value ANOVA p value; HR [95% CI] p valuee ANOVA p value; HR [95% CI] p value Rag2 –/–/Il2rg –/– Rag2 –/–/Il2rg –/– Rag2 –/–/Il2rg –/– n = 23 p = 0.0011; 3.32 [1.43; 7.68] p = 0.0051 n = 23 p = 0.0009; 2.94 [1.29; 6.71] p = 0.0104 n = 21 p = 0.9045; 1.00 [0.48; 2.07] p = 0.9978 α-NKG2D α-NKG2D α-NKG2D n = 19 p = 0.0018; 3.23 [1.63; 6.40] p = 0.0008 n = 19 p = 0.0603; 2.27 [1.17; 4.41] p = 0.0147 n = 19 p = 0.1870; 0.97 [0.53; 1.79] p = 0.9286 Il22 –/– Il22 –/– Il22 –/– n = 8 p = 0.0437; 2.22 [0.84; 5.87] p = 0.1076 n = 8 p = 0.5761; 1.25 [0.48; 3.24] p = 0.6513 n = 8 p = 0.0863; 0.55 [0.21; 1.41] p = 0.1495 –/– Ifng –/– –/– Ifng p = 0.0508; 5.28 [1.87; 14.92] p = 0.0017 p = 0.0648; 3.68 [1.38; 9.82] p = 0.0094 Ifng p = 0.9703; 1.14 [0.45; 2.87] p = 0.7362 n = 9 n = 10 n = 9 –/– Il21 –/– –/– Il21 p = 0.0286; 2.15 [1.06; 4.33] p = 0.0329 p = 0.0058; 2.52 [1.24; 5.09] p = 0.0102 Il21 p = 0.3151; 1.52 [0.78; 2.95] p = 0.1849 n = 20 n = 21 n = 20 –/– –/– –/– Myd88 p = 0.1693; 1.88 [0.66; 5.37] p = 0.2374 Myd88 p = 0.3203; 2.19 [0.73; 6.54] p = 0.1617 Myd88 p = 0.9539; 0.91 [0.33; 2.47] p = 0.8333 n = 6 n = 6 n = 6 –/– –/– –/– Il21r p = 0.0838; 1.54 [0.71; 3.35] p = 0.2743 Il21r p = 0.1624; 1.32 [0.61; 2.88] p = 0.4778 Il21r p = 0.6151; 0.86 [0.40; 1.86] p = 0.6820 n = 13 n = 12 n = 12 α-IFNAR1 α-IFNAR1 α-IFNAR1 n = 21 p = 0.0570; 2.24 [1.19; 4.22] p = 0.0123 n = 21 p = 0.3814; 1.60 [0.84; 3.07] p = 0.1539 n = 21 p = 0.8238; 0.88 [0.49; 1.60] p = 0.6596 α-IFNγ α-IFNγ α-IFNγ n = 23 p = 0.4145; 1.19 [0.65; 2.16] p = 0.5751 n = 22 p = 0.2869; 1.50 [0.84; 2.71] p = 0.1739 n = 22 p = 0.0530; 1.71 [0.93; 3.13] p = 0.0418 α α α α-CD8α+αIL17 α α α -CD8 + IL17 p = 0.3869; 1.65 [0.66; 4.12] p = 0.2831 p = 0.4415; 1.54 [0.62; 3.87] p = 0.3345 -CD8 + IL17 p = 0.8364; 0.91 [0.37; 2.25] p = 0.8270 n = 10 n = 10 n = 10 α α α α-CD4+α-CD8α α α α -CD4+ -CD8 p = 0.3212; 0.56 [0.24; 1.31] p = 0.1846 p = 0.4776; 0.61 [0.26; 1.40] p = 0.2402 -CD4+ -CD8 p = 0.9881; 1.09 [0.46; 2.56] p = 0.8414 n = 12 n = 12 n = 10 α α-CD4 α -CD4 p = 0.8438; 1.02 [0.48; 2.18] p = 0.9561 p = 0.7883; 1.09 [0.52; 2.28] p = 0.8230 -CD4 p = 0.2510; 1.42 [0.61; 3.32] p = 0.3786 n = 18 n = 18 n = 11 –/– –/– –/– Nlrp3 p = 0.6765; 1.52 [0.58; 4.01] p = 0.3992 Nlrp3 p = 0.6943; 1.41 [0.54; 3.66] p = 0.4802 Nlrp3 p = 0.8508; 0.89 [0.34; 2.31] p = 0.7828 n = 8 n = 8 n = 8 α α α -IL17 p = 0.8442; 0.93 [0.36; 2.38] p = 0.8782 -IL17 p = 0.8963; 0.96 [0.38; 2.42] p = 0.9303 -IL17 p = 0.1772; 0.50 [0.19; 1.34] p = 0.0954 n = 17 n = 17 n = 9 Control Control Control n = 114 n = 114 n = 109 ANAK p = 0.6379; 0.78 [0.45; 1.36] p = 0.3821 ANAK p = 0.8682; 0.86 [0.50; 1.49] p = 0.5923 ANAK p = 0.7326; 1.06 [0.64; 1.76] p = 0.8016 n = 30 n = 31 n = 29 –/– –/– –/– Casp1 p = 0.9619; 1.01 [0.40; 2.51] p = 0.9896 Casp1 p = 0.6736; 1.27 [0.50; 3.21] p = 0.6141 Casp1 p = 0.9669; 1.11 [0.41; 3.00] p = 0.8037 n = 7 n = 7 n = 7 α α α α α α -CD8 +ANAK p = 0.9923; 1.05 [0.41; 2.68] p = 0.9173 -CD8 +ANAK p = 0.6073; 0.84 [0.33; 2.14] p = 0.7165 -CD8 +ANAK p = 0.7298; 1.18 [0.48; 2.90] p = 0.6772 n = 10 n = 10 n = 10 –/– –/– Rag2 p = 0.6909; 0.83 [0.31; 2.25] p = 0.7187 Rag2 p = 0.4890; 0.93 [0.34; 2.54] p = 0.8942 Rag2 –/– p = 0.3382; 0.56 [0.21; 1.46] p = 0.1933 n = 7 n = 10 n = 7 –/– Il1r1 –/– p = 0.9590; 1.00 [0.40; 2.53] p = 0.9995 Il1r1 p = 0.7956; 1.19 [0.47; 3.02] p = 0.7086 Il1r1 –/– p= 0.9669; 1.61 [0.62; 4.17] p = 0.1898 n = 9 n = 9 n = 9 α α α α α α -CD8 p = 0.1753; 0.76 [0.38; 1.50] p = 0.4280 -CD8 p = 0.0966; 0.55 [0.28; 1.10] p = 0.0902 -CD8 p = 0.1074; 0.71 [0.37; 1.37] p = 0.2627 n = 27 n = 28 n = 17

–100 –50 0 50 100 –100 –50 0 50100 150 200 –60 –30 0 3060 90 120 Δ TFS (Δ days vs. control average) OS (Δ days vs. control average) TTD ( days vs. control average)

f WT (TFS, n = 10; OS, n = 10) g Control (TFS, n = 25; OS, n = 25) h Control (TFS, n = 12; OS, n = 12) –/– Rag2 (TFS, n = 7; OS, n = 10) α-NKG2D (TFS, n = 19; OS, n = 19) α-CD4+α-CD8α (TFS, n = 12; OS, n = 12) 100 HR: 0.83 [0.31; 2.25] 100 HR: 0.93 [0.34; 2.54] 100 HR: 3.23 [1.63; 6.40] 100 HR: 2.27 [1.17; 4.41] 100 HR: 0.56 [0.24; 1.31] 100 HR: 0.61 [0.26; 1.40] p = 0.7187 p = 0.8942 p = 0.0008 p = 0.0147 p = 0.1846 p = 0.2402 75 75 75 75 75 75

50 50 50 50 50 50 OS (%) OS (%) OS (%) TFS (%) TFS (%) 25 25 TFS (%) 25 25 25 25

0 0 0 0 0 0 50 100 150 50 100 150 200 40 80 120 160 50 100 150 200 40 80 120 160 200 40 80120 160 200 Days since 1st gavage Days since 1st gavage Days since 1st gavage Days since 1st gavage Days since 1st gavage Days since 1st gavage

Fig. 2 M/D-driven oncogenesis is under immunosurveillance by NK cells. a, b Tumor-free survival (TFS), overall survival (OS) and time-to-death (TTD) of WT C57BL/6 mice (a, b) and Rag2−/−Il2rg−/− mice (a)orIfng−/− mice (b) subjected to M/D-driven oncogenesis. Number of mice, hazard ratio (HR) and p values (two-sided log-rank) are reported. c–e. Variations in TFS (c), OS (d) and TTD (e) imposed to M/D-driven oncogenesis in mice by the indicated genotype or immunomodulatory interventions. Results are means ± SEM plus individual data points. Number of mice, HR and p values (two-sided log-rank and one way-ANOVA plus Fisher LSD, calculated with respect to individual control experiments) are reported. Green dots indicate mice that were free of disease (c) and alive (d) at the end of the experiment. f–h TFS and OS of WT C57BL/6 (f–h) mice and Rag2−/− BALB/c mice (f), C57BL/6 mice receiving NKG2D-depleting antibodies (g), or C57BL/6 mice receiving CD4- and CD8-depleting antibodies (h). Number of mice, HR and p values (two- sided log-rank) are reported. oncogenesis in this setting (Fig. 2c). These effects were also Importantly, while M/D-driven oncogenesis was fostered by a manifest at the level of OS for the Ifng−/− and Il21−/− genotypes severe immunodeficiency affecting T lymphocytes, B cells and NK and for NKG2D+ cell depletion (Fig. 2d). Disease progression cells (Rag2−/−Il2rg−/− genotype), TFS and OS in Rag2−/− mice computed as time from detection to death (time to death, TTD) (who lack T lymphocytes and B cells, but have an intact NK cell was only accelerated by IFNG-neutralizing antibodies (Fig. 2e). compartment)9 did not differ from TFS and OS in WT mice

4 NATURE COMMUNICATIONS | (2020) 11:3819 | https://doi.org/10.1038/s41467-020-17644-0 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-17644-0 ARTICLE

(Fig. 2f). Moreover, TFS and OS were negatively affected by the cause a stable reduction in body mass as compared to unrestricted depletion of NKG2D+ cells (encompassing NK cells and T cells), access to standard chow (Supplementary Fig. 1g). Along similar but not by the co-depletion of CD4+ and CD8+ lymphocytes lines, dietary supplementation with NAM (Fig. 4b) and nicoti- (Fig. 2g, h). Altogether, these findings establish the importance of namide riboside (NR), which are two variants of vitamin B , + 3 immunoevasion for the emergence and progression of HR BC, pyridoxine (vitamin B6) and calcitriol (vitamin D3) all tended to placing emphasis on interferon signaling in the context of NK ameliorate TFS and OS in WT C57BL/6 subjected to M/D-driven cell-mediated immunosurveillance. In further support of this carcinogenesis (Fig. 4c–e). In contrast, dietary supplementation notion, the partial oncopreventive effect of tamoxifen was lost in with spermidine (a polyamine with prominent life-extending Rag2−/−Il2rg−/− mice (Supplementary Fig. 1f). effects)19 failed to affect M/D-driven oncogenesis (Fig. 4c–e). Amongst all vitamins tested, NAM delivered with the drinking M/D-driven tumors are poorly immunoedited by T cells.To water (which was more active than the administration of NR with further explore immunosurveillance in HR+ BC, we established chow) was the only one to completely (>300 days) block onco- genesis in a considerable fraction (~25%) of mice, and to robustly cell lines from M/D-driven tumors developing in immuno- – competent mice. When injected subcutaneously (s.c.) or into the delay disease progression (Fig. 4a e). Of note, dietary supple- mammary fat pad of immunodeficient Rag2−/− mice, these cell mentation of NAM to C57BL/6 mice resulted in circulating NAM levels of ~16.3 µg mL−1 µM (Supplementary Fig. 1h), while NAM lines rapidly generated cancers with aggressive, MPA-insensitive −1 proliferation (Fig. 3a, b). In sharp contrast, these cell lines only concentrations around 2.3 µg mL have been documented in sporadically (in 2 out of 52 attempts) generated progressive healthy volunteers receiving 2 g nicotinic acid (a precursor of 20 fi lesions upon s.c. injection into immunocompetent, syngeneic NAM) per os . The bene cial effects of NAM could also be C57BL/6 hosts (Fig. 3c), and invariably failed to establish pro- documented in a transgenic model of immunoevasive, luminal B BC driven by the polyomavirus middle T (PyMT) antigen gressing neoplasms upon orthotopic implantation (Fig. 3d), 21,22 irrespective of whether mice received MPA. Often, small tumors expressed under control of the MMTV promoter (Supple- (generally <50 mm2 surface) formed, but spontaneously regressed mentary Fig. 1i). over time (Fig. 3a–d). DMBA administration, alone or in com- A high sucrose diet (HSD) accelerated M/D-driven oncogen- bination with MPA, failed to promote the growth of M/D-driven esis, an effect that was completely abolished by concomitant cancer cells injected s.c. in syngeneic immunocompetent reci- NAM supplementation, which continued to extend TFS and OS pients, although it caused the development of endogenous tumors beyond the level of control mice fed a normal chow (although + only one mice remained cancer-free) (Fig. 4c–e). Less prominent (in 5/20 mice with DMBA alone, and 20/20 mice with DMBA fi MPA) (Fig. 3e), implying that DMBA does not cause severe bene cial effects on TFS, OS or TTD were observed when HSD fi was provided in the context of pyridoxine or calcitriol immunode ciency in this model. As an additional control, – mammary cancer cell lines established from M/D-driven tumors supplementation (Fig. 4c e). A dietary regimen rich in fat −/− (high-fat diet, HFD) accelerated M/D-driven carcinogenesis and evolving in Rag2 mice always failed to develop tumors when – transplanted into WT mice (Fig. 3f). shortened OS, but had no impact on TTD (Fig. 4c e). Such a detrimental effect of HFD was fully antagonized by NAM but not by calcitriol or pyridoxine (Fig. 4c–e). In this context, NAM also Prophylactic vaccination delays M/D-driven carcinogenesis. reversed the increase in body weight caused by HFD (Supple- We next investigated whether vaccinating immunocompetent mentary Fig. 1f). In conclusion, M/D-driven tumorigenesis is mice with cell lines established from M/D-driven cancers would sensitive to dietary alterations with an established detrimental have a preventive effect on M/D-driven oncogenesis. To this aim, effect on human BC, including HSD and HFD, and NAM cell lines established from M/D-driven tumors evolving in mediates robust oncopreventive effects in all these scenarios. Rag2−/− or WT mice were killed in vitro with the immunogenic cell death inducer mitoxantrone (MTX)15,16 and injected into the fat pad of WT mice 1 week prior to, in the 4th week of, and 2 days NAM mediates chemopreventive effects that depend on T cells. after de novo M/D-driven carcinogenesis. This treatment post- We observed that the oncopreventive effects of NAM, pyridoxine poned the manifestation of M/D-driven tumors irrespective of the −/− −/− −/− and calcitriol were invariably lost in Rag2 Il2rg mice immunological status of the host (Rag2 vs WT) from which (Fig. 5a–c). Moreover, NAM−dependent oncoprevention was cell lines were derived (Fig. 3g). Importantly, injection of orga- −/− — abolished in Rag2 mice, in mice subjected to the antibody- noids from the normal breast epithelium optionally treated with mediated co-depletion of CD4+ and CD8+ T cells, in mice MTX or radiation therapy (another inducer of immunogenic cell receiving an IFNG-neutralizing antibody, as well as in Ifnar1−/− 15,16— death) had a similar capacity to delay M/D-driven onco- mice (Fig. 5d–g). These data lend further support to the notion genesis in C57BL/6 mice (Fig. 3h, i). In this setting, neutralization — that interferon signaling is key for the immunosurveillance of of annexin A1 (ANXA1) a danger signal involved in the per- HR+ BC. Moreover, they demonstrate that nutritional interven- 15,16— ception of cell death as immunogenic abolished the vacci- tions with oncopreventive effects in this model, such as NAM nation effect (Fig. 3i). Thus, immunosurveillance of M/D-driven supplementation, boost T cell-dependent immunosurveillance. carcinogenesis can be boosted, at least to some degree, by NAM has been suggested to trigger autophagy, a cytoprotective immunological interventions including vaccination. pathway involving the lysosomal degradation of disposable or potentially dangerous cytosolic entities23. However, we failed to Safe nutritional measures delay M/D-driven carcinogenesis. detect consistent biochemical signs of autophagy, such as the Multiple epidemiological studies indicate that nutritional status lipidation of microtubule associated protein 1 light chain 3 beta has a major impact on BC incidence17, and it is now established (MAP1LC3B, best known as LC3) and the degradation of that dietary interventions can mediate therapeutically relevant sequestosome 1 (SQSTM1, best known as p62) in M/D-driven immunostimulatory effects18. We therefore analyzed the impact tumors developing despite NAM administration, as well as in of nutritional interventions on M/D-driven carcinogenesis in WT multiple healthy tissues (Supplementary Fig. 2a). Moreover, the C57BL/6 mice. We found that biweekly 24-h-long fasting cycles oncopreventive effects of NAM were conserved in Becn1+/− mice postponed M/D-driven oncogenesis, decelerated disease pro- (which exhibit partial autophagic defects and increased suscept- gression and hence extended OS (Fig. 4a), although they did not ibility to mammary carcinogenesis)24,25 (Supplementary Fig. 2b).

NATURE COMMUNICATIONS | (2020) 11:3819 | https://doi.org/10.1038/s41467-020-17644-0 | www.nature.com/naturecommunications 5 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-17644-0

Subcutaneous grafts into Rag2 –/– mice Orthotopic grafts into Rag2 –/– mice

a Control 16/16 +MPA 15/15 b Control 15/15 +MPA 15/15 2.5 2.5 2.5 2.5 2.0 2.0 2.0 2.0 ) ) ) ) 2 2 2 2 1.5 1.5 1.5 1.5 1.0 1.0 1.0 1.0 Area (cm Area (cm Area (cm 0.5 0.5 Area (cm 0.5 0.5 0.0 0.0 0.0 0.0 0 25 50 75 100 0 25 50 75 100 0 25 50 75 100 0 25 50 75 100 Days after injection Days after injection Days after injection Days after injection Subcutaneous grafts into WT mice Orthotopic grafts into WT mice

c Control 1/36 +MPA 1/16 d Control 0/16 +MPA 0/16 2.5 2.5 2.5 2.5 2.0 2.0 2.0 2.0 ) ) ) ) 2 2 2 2 1.5 1.5 1.5 1.5 1.0 1.0 1.0 1.0 Area (cm Area (cm Area (cm 0.5 0.5 Area (cm 0.5 0.5 0.0 0.0 0.0 0.0 0 25 50 75 100 02550 75 100 0 25 50 75 100 02550 75 100 Days after injection Days after injection Days after injection Days after injection Subcutaneous grafts into WT mice

e +DMBA 0/20 +MPA/DMBA 0/20 f Control 0/15 Endogenous Endogenous 2.5 2.5 2.5 lD MGT_1 2.0 2.0 lC MGT_1 Subcutaneous 2.0 lD MGT_2 ) ) ) 2 2 2 1.5 1.5 lC MGT_2 tumors 1.5 lD MGT_3 1.0 1.0 lC MGT_3 1.0 S.C. S.C.

Area (cm lC MGT_4 Area (cm 0.5 0.5 Area (cm 0.5 Endogenous DMBA- 0.0 0.0 or M/D-driven tumors 0.0 0 50 100 150 200 250 0 50 100 150 200 250 02550 75 100 Days since 1st gavage Days since 1st gavage Days after injection No vaccination (n = 8) g 100 100 100 IC MGT + MTX (n = 12) 75 75 75 ID MGT + MTX (n = 7)

p = 0.0024 50 50 50 HR: 0.31 [0.09;0.98] TFS p = 0.0734 HR: 0.44 [0.15;1.30] OS (%) TFS (%) 25 25 TTD (%) 25 p = 0.4162 HR: 0.71 [0.28;1.81] OS p = 0.2330 HR: 0.56 [0.19;1.61] 0 0 0 p = 0.3917 HR: 1.45 [0.59;3.58] 80 120 160 200 240 40 80 120 160 200 0 306090120TTD p = 0.9585 HR: 1.79 [0.62;5.18] Days since 1st gavage Days since 1st gavage Days after detection No vaccination (n = 17) h 100 100 100 MMGO (n = 16) 75 75 75 MMGOs + 20Gy (n = 10) p = 0.0136 HR: 0.47 [0.28;0.97] 50 50 50 TFS p = 0.0059 HR: 0.34 [0.15;0.77] OS (%) TFS (%) 25 25 TTD (%) 25 p = 0.1027 HR: 0.60 [0.30;1.21] OS p = 0.0160 HR: 0.41 [0.19;0.90] 0 0 0 p = 0.1298 HR: 1.62 [0.80;3.30] 40 100 160 220 280 60 120 180 240 300 0 255075100TTD p = 0.3322 HR: 1.50 [0.57;3.97] Days since 1st gavage Days since 1st gavage Days after detection No vaccination (n = 17) i 100 100 100 MMGOs + MTX (n = 16) 75 75 75 MMGOs + MTX + α-ANXA1 (n = 10)

50 50 50 p = 0.0033 HR: 0.39 [0.19;0.84] TFS p = 0.9716 HR: 0.98 [0.45;2.15] OS (%) TFS (%) 25 25 TTD (%) 25 p = 0.0943 HR: 0.58 [0.29;1.19] OS p = 0.9826 HR: 0.99 [0.45;2.16] 0 0 0 p = 0.7380 HR: 0.89 [0.44;1.81] 40 100 160 220 280 60 120 180 240 300 0 50 100 150 200 TTD p = 0.9487 HR: 1.02 [0.47;2.24] Days since 1st gavage Days since 1st gavage Days after detection

Similarly, M/D-driven oncogenesis remained sensitive to NAM- node cells. In this setting, we detected early upregulation of gene mediated oncoprevention in mice subjected to the conditional sets involved in immune functions, notably type I interferon knockout of Atg5 (a key component of the autophagy machinery)26 signaling and chemotaxis (Fig. 5h, i). However, we were unable to in keratin 5 (KRT5)-expressing cells (including cells of the detect consistent quantitative or proliferative alterations in the 27 + + + + mammary epithelium) (Supplementary Fig. 2c), indicating that CD8 T cell, CD4 T cell, CD4 FOXP3 regulatory T (TREG), NK this effect does not require autophagy induction in malignant cells. cell, NKT cell, and B cell compartments of the spleen and lymph To further explore the immunostimulatory activity of NAM, nodes of tumor-naïve C57BL/6 mice receiving NAM in the tumor-free C57BL/6 mice were treated with NAM for 1, 3 or drinking water for 14 days (Supplementary Fig. 2d, e), suggesting 14 days, followed by RNAseq analyses of splenocytes and lymph that the ability of NAM to delay M/D-driven carcinogenesis

6 NATURE COMMUNICATIONS | (2020) 11:3819 | https://doi.org/10.1038/s41467-020-17644-0 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-17644-0 ARTICLE

Fig. 3 M/D-driven tumors are poorly immunoedited by T cells and sensitive to vaccination. a–d Growth of cell lines established from M/D-driven tumors (herein called mammary gland tumors, MGTs) developing in immunocompetent (IC) mice upon subcutaneous (a, c) or orthotopic (b, d) implantation in Rag2−/− (a, b) or IC C57BL/6 mice (c, d) optionally bearing an MPA-releasing pellet. Individual curves and tumor incidence are reported. e Growth of MGTs upon subcutaneous implantation in IC C57BL/6 mice receiving DMBA alone or in the context of an MPA-releasing pellet. Individual curves for all tumors and subcutaneous tumor incidence are reported. Boxed curves refer to endogenous tumors caused by DMBA alone or MPA plus DMBA. f Growth of MGTs established from M/D-driven tumors developing in immunodeficient (ID) Rag2−/− mice upon subcutaneous implantation in IC C57BL/6 mice. Individual curves and global tumor incidence are reported. g Tumor-free survival (TFS), overall survival (OS) and time-to-death (TTD) of WT C57BL/6 mice subjected to M/D-driven oncogenesis in control conditions or upon vaccination with cell lines established from M/D-driven tumors developing in IC or ID mice and treated with mitoxantrone (MTX) in vitro. Number of mice, hazard ratio (HR) and p values (two-sided log-rank, as compared to non-vaccinated mice) are reported. h TFS, OS, and TTD of WT C57BL/6 mice subjected to M/D-driven oncogenesis in control conditions or upon vaccination with mouse mammary gland organoids (MMGOs) optionally killed in vitro by exposure to radiation therapy (RT) in a single dose of 20 Gy. Number of mice, HR and p values (two-sided log-rank, as compared to non-vaccinated mice) are reported. Please note that part of these results (control conditions) are also depicted in (i). i TFS, OS, and TTD of WT C57BL/6 mice subjected to M/D-driven oncogenesis in control conditions or upon vaccination with MMGOs killed in vitro by mitoxantrone (MTX) administration in the optional context of ANXA1 neutralization. Number of mice, HR and p values (two-sided log-rank, as compared to non-vaccinated mice) are reported. Please note that part of these results (control conditions) are also depicted in (h).

a Control (TFS, n = 20; OS, n = 20; TTD, n = 19) b Control (TFS, n = 30; OS, n = 30; TTD, n = 29) Fasting (TFS, n = 20; OS, n = 20; TTD, n = 16) NAM (TFS, n = 28; OS, n = 28; TTD, n = 17) 100 100 100 100 100 100 HR: 0.34 [0.17;0.70] HR: 0.36 [0.18;0.73] HR: 0.63 [0.32;1.22] HR: 0.32 [0.17;0.57] HR: 0.32 [0.18;0.57] HR: 0.58 [0.33;1.04] p = 0.0002 p = 0.0005 p = 0.1218 p<0.0001 p<0.0001 p = 0.0511 75 75 75 75 75 75

50 50 50 50 50 50 TTD (%) OS (%) OS (%) TFS (%) TTD (%) 25 25 TTD (%) 25 25 25 25

0 0 0 0 0 0 50 100 150 200 250 300 50 100 150 200 250 300 02040 60 80 100 50 100 150 200 250 300 50 100 150 200 250 300 0 20406080 Days since 1st gavage Days since 1st gavage Days after detection Days since 1st gavage Days since 1st gavage Days after detection

cdeANOVA p value; HR [95% Cl] p value ANOVA p value; HR [95% Cl] p value ANOVA p value; HR [95% Cl] p value

HFD + Pyridoxine p = 0.0043; 4.54 [1.45; 14.19] p<0.0001 HFD + Pyridoxine p = 0.0067; 4.30 [1.40; 13.18] p<0.0001 HFD + Pyridoxine p = 0.6128; 1.51 [0.61; 3.75] p = 0.3198 n = 10 n = 10 n = 10 HFD + Calcitriol p = 0.0057; 4.69 [1.48; 14.85] p<0.0001 HFD + Calcitriol p = 0.0178; 3.68 [1.26; 10.74] p = 0.0004 HFD + Calcitriol p = 0.7698; 0.95 [0.39; 2.33] p = 0.3198 n = 10 n = 10 n = 10 HSD p = 0.0086; 4.56 [1.46; 14.26] p<0.0001 HSD p = 0.1031; 1.61 [0.65; 3.98] p = 0.2235 HSD p = 0.0318; 0.54 [0.19; 1.55] p = 0.2233 n = 10 n = 10 n = 10 HFD p = 0.0090; 2.67 [1.30; 5.51] p = 0.0028 HFD p = 0.0270; 2.33 [1.14; 4.78] p = 0.0112 HFD p = 0.9835; 1.13 [0.52; 2.45] p = 0.7326 n = 20 n = 20 n = 20 HSD + Calcitriol p = 0.1060; 1.87 [0.74; 4.68] p = 0.1165 HSD + Calcitriol p = 0.7226; 0.79 [0.33; 1.90] p = 0.5624 HSD + Calcitriol p<0.0001; 0.28 [0.09; 0.86] p = 0.0009 n = 10 n = 10 n = 10 NAM + HFD p = 0.3698; 1.41 [0.67; 2.94] p = 0.3555 NAM + HFD p = 0.6340; 1.32 [0.63; 2.77] p = 0.4436 NAM + HFD p = 0.3381; 0.78 [0.34; 1.82] p = 0.5256 n = 19 n = 19 n = 17 Control Control Control n = 30 n = 30 n = 29 Spermidine p = 0.8863; 0.99 [0.41; 2.37] p = 0.9760 Spermidine p = 0.8220; 1.22 [0.51; 2.95] p = 0.6244 Spermidine p = 0.4735; 1.73 [0.70; 4.30] p = 0.1691 n = 10 n = 10 n = 10 HSD + Pyridoxine p = 0.8797; 1.30 [0.50; 3.37] p = 0.5673 HSD + Pyridoxine p = 0.6242; 1.13 [0.44; 2.89] p = 0.7871 HSD + Pyridoxine p = 0.1207; 0.54 [0.19; 1.51] p = 0.2233 n = 10 n = 10 n = 5 HSD + NAM p = 0.0327; 0.37 [0.14; 0.98] p = 0.0086 HSD + NAM p = 0.0065; 2.35 [0.13; 0.95] p = 0.0066 HSD + NAM p = 0.0003; 0.37 [0.13; 1.06] p = 0.0188 n = 10 n = 10 n = 6 NR p = 0.0477; 0.45 [0.17; 1.17] p = 0.0465 NR p = 0.0775; 0.47 [0.18; 1.21] p = 0.0738 NR p = 0.8946; 0.76 [0.28; 2.05] p = 0.5549 n = 10 n = 10 n = 7 Pyridoxine p = 0.0191; 0.36 [0.13; 0.98] p = 0.0114 Pyridoxine Pyridoxine p = 0.5606; 0.63 [0.25; 1.58] p = 0.2792 n = 10 n = 10 p = 0.0010; 0.33 [0.12; 0.90] p = 0.0040 n = 9 Calcitriol p = 0.0225; 0.40 [0.15; 1.06] p = 0.0194 Calcitriol Calcitriol p = 0.1142; 0.39 [0.14; 1.05] p = 0.0280 n = 10 n = 10 p = 0.0050; 0.42 [0.16; 1.10] p = 0.0309 n = 7 Fasting p = 0.0002; 0.34 [0.17; 1.70] p = 0.0002 Fasting Fasting p = 0.0096; 0.63 [0.32; 1.22] p = 0.1218 n = 20 n = 20 p<0.0001; 0.36 [0.18; 0.73] p = 0.0005 n = 16 NAM NAM NAM n = 28 p<0.0001; 0.32 [0.17; 0.57] p<0.0001 n = 28 p<0.0001; 0.32 [0.18; 0.57] p = 0.0001 n = 17 p = 0.1738; 0.58 [0.33; 1.04] p = 0.0511

–100 0 100 200 –100 0 100 200 –100 0 100 200 TFS (Δ days vs. control average) OS (Δ days vs. control average) OS (Δ days vs. control average)

Fig. 4 Safe nutritional interventions delay M/D-driven oncogenesis. a, b Tumor-free survival (TFS), overall survival (OS) and time-to-death (TTD) of C57BL/6 mice subjected to M/D-driven oncogenesis in control conditions (a, b) or in the context of weekly 24-h-fasting episodes (a) or NAM supplementation with the drinking water (b). Number of mice, hazard ratio (HR) and p values (two-sided log-rank) are reported. c–e Variations in TFS (c), OS (d) and TTD (e) imposed to M/D-driven oncogenesis in C57BL/6 mice by the indicated nutritional interventions. Results are means ± SEM plus individual data points. Number of mice, HR and p values (two-sided log-rank and one way-ANOVA plus Fisher LSD, calculated with respect to individual control experiments) are reported. Green dots indicate mice that were free of disease (c) and alive (d) at the end of the experiment. HFD high-fat diet, HSD high sucrose diet, NR nicotinamide riboside. involves immunological pathways not linked to the systemic TSA cells)28,29 as well as in transplantable fibrosarcoma sarcoma expansion/contraction of immune effectors. In summary, NAM models (MCA205 cells) established in immunocompetent syn- appears to mediate oncopreventive effects through an autophagy- geneic hosts (Supplementary Fig. 3b–d). This effect did not independent, immunological mechanism that relies on lymphoid depend on autophagy activation, as demonstrated in AT3 cells cells (T cells and NK cells) as well as interferon signaling. subjected to the shRNA-dependent depletion of ATG7 (a core component of the autophagy machinery)30, as well as in Atg7−/− NAM mediates immunotherapeutic activity against BC. Con- TSA cells (Supplementary Fig. 3e, f). tinuous NAM supplementation delayed oncogenesis and hence To mechanistically explore the therapeutic effects of NAM, we prolonged OS in mice exposed to M/D-driven carcinogenesis compared the transcriptomic profile of untreated vs NAM-treated (Fig. 4b–e). When dietary NAM supplementation was started at M/D-driven tumors, revealing signs of immune regulation (rather than before) tumor detection, NAM-mediated therapeutic (Fig. 6c). In particular, NAM administration restored some of effects that were more pronounced than those observed with the physiological transcriptional features that were lost in the continuous NAM provision from oncogenesis (Fig. 6a, b), and course of M/D-driven oncogenesis (Fig. 1h), such as the were not associated with the loss of the luminal B (ER+VIM−) expression of genes associated with germinal centers (Cd22, phenotype (Supplementary Fig. 3a). NAM-mediated therapeutic Dock10, Gimap8) (Fig. 6c), and—to a lesser degree—lymphoid activity also in transplantable models of luminal B BC (AT3 and cell activation (Cd48, fold change = 2.23, p = 1.62E−02; Cd160,

NATURE COMMUNICATIONS | (2020) 11:3819 | https://doi.org/10.1038/s41467-020-17644-0 | www.nature.com/naturecommunications 7 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-17644-0

a Rag2 –/–/ll2rg –/– (n = 11) +NAM (n = 11) b Rag2 –/–ll2rg –/– (n = 12) +Pyridoxine (n = 12) 100 100 100 100 HR: 0.66 [0.29; 1.67] HR: 0.49 [0.20; 1.19] HR: 1.72 [0.74; 3.98] HR: 1.57 [0.68; 3.59] p = 0.3189 p = 0.0707 p = 0.1425 p = 0.2388 75 75 75 75

50 50 50 50 OS (%) OS (%) TFS (%) TFS (%) 25 25 25 25

0 0 0 0 40 60 80 100 120 140 40 60 80 100 120140 160 180 40 60 80 100 120 140 40 60 80 100 120140 160 Days since 1st gavage Days since 1st gavage Days since 1st gavage Days since 1st gavage

c Rag2 –/–/ll2rg –/– (n = 12) +Calcitriol (n = 12) d Rag2 –/– (n = 10) +NAM (n = 10) 100 HR: 0.91 [0.40; 2.07] 100 HR: 0.69 [0.29; 1.67] 100 HR: 0.55 [0.18; 1.64] 100 HR: 0.63 [0.26; 1.56] p = 0.8098 p = 0.3953 p = 0.3391 p = 0.2672 75 75 75 75

50 50 50 50 OS (%) OS (%) TFS (%) TFS (%) 25 25 25 25

0 0 0 0 40 60 80 100 120 140 40 60 80 100 120 140 160 40 80 120 160 200 50 100 150 200 Days since 1st gavage Days since 1st gavage Days since 1st gavage Days since 1st gavage

e α-CD4+α-CD8α (n = 12) +NAM (n = 12) f α-IFNγ (n = 12) +NAM (n = 12) 100 HR: 0.87 [0.38; 2.01] 100 HR: 1.28 [0.50; 2.56] 100 HR: 1.62 [0.71; 3.69] 100 HR: 1.13 [0.51; 2.52] p = 0.7358 p = 0.7616 p = 0.1964 p = 0.7483 75 75 75 75

50 50 50 50 OS (%) OS (%) TFS (%) TFS (%) 25 25 25 25

0 0 0 0 40 80 120 160 200 240 280 40 80 120160 200 240 280 40 80 120 160 200 240 280 40 80 120 160 200 240 280 Days since 1st gavage Days since 1st gavage Days since 1st gavage Days since 1st gavage

–/– ghilfnar1 (n = 9) +NAM (n = 12) Spleen Lymph node 0.8 30 100 HR: 0.82 [0.34; 2.00] 100 HR: 0.84 [0.32; 2.03] p = 0.6269 p = 0.6720 75 0.6 75 20

50 50 0.4 Expression OS (%) TFS (%) Expression 10 25 25 0.2

0 0 0 0 50 75 100 125 150 50 75 100 125150 175 Control 1 d 3 d 14 d Control 1 d 3 d 14 d Days since 1st gavage Days since 1st gavage Cluster 1: 423 genes Cluster 1: 102 genes Oxidation-reduction process 4.2E–15 Complement activation 9.7E–05 Metabolic process 7.9E–15 B-cell receptor signaling pathway 9.7E–05 Lipid metabolic process 2.6E–09 Regulation of B-cell activation 3.9E–04 Cluster 2: 71 genes Cluster 2: 52 genes Response to virus 1.1E–08 Immune system process 4.1E–05 Cellular response to IFNb 1.0E–05 Innate immune response 4.4E–04 Defense response to virus 1.9E–05 Regulation on lL-4 production 1.0E–03 Cluster 3: 85 genes Cluster 3: 68 genes Chemokine-mediated signalling 3.3E–03 Regulation of pH 1.8E–03 Neutrophil chemotaxis 5.1E–03 Inflammatory response 8.1E–03 Immune response 8.4E–03 Regulation on phosphatase activity 1.6E–02

Fig. 5 NAM mediates chemopreventive effects by boosting immunosurveillance. a–c Tumor-free survival (TFS) and overall survival (OS) of Rag2−/− Il2rg−/− mice subjected to M/D-driven oncogenesis in control conditions (a–c) or along with NAM (a), pyridoxine (b) or calcitriol (c) supplementation. Number of mice, hazard ratio (HR) and p values (two-sided log-rank) are reported. d–g TFS and OS of Rag2−/− mice (d), C57BL/6 mice receiving CD4- and CD8α-depleting antibodies (e), C57BL/6 mice receiving an IFNG-neutralizing antibody (f), and Ifnar1−/− mice (g) subjected to M/D-driven oncogenesis in control conditions or along with NAM supplementation with the drinking water. Number of mice, HR and p values (two-sided log-rank) are reported. h, i Gene sets enriched in the spleen (h) and lymph nodes (i) of tumor-naive C57BL/6 mice receiving NAM supplementation with the drinking water for 1, 3 or 14 days. Adjusted, two-sided p values for enrichment are reported. fold change = 2.37, p = 4.23E−02; Cd244, fold change = 2.84, counterparts (at a similar size), we failed to identify consistent = − + + p 5.07E 03). A similar study of untreated vs NAM-treated AT3 changes in the abundance of CD8 , CD4 T cells and TREG cells, tumors identified transcriptional patterns linked to T-cell NK and NKT cells, and B cells (Supplementary Fig. 4a). differentiation, as well as to type I and type III interferon Moreover, we failed to detect significant differences on CD8+, + signaling (Supplementary Fig. 3g). Only 23 genes were con- CD4 T cells and TREG cells from untreated vs NAM-treated sistently upregulated in both M/D-driven and AT3 mammary tumors with respect to proliferation (Ki67 positivity) (Supple- tumors responding to NAM (Fig. 6d and Supplementary Fig. 3g). mentary Fig. 4a). Similar observations, which we attribute to the Ingenuity pathway analysis indicated that most of these genes considerable heterogeneity of the model, were made in the spleen positively regulate immunological processes including the and lymph nodes of C57BL/6 mice bearing M/D-driven tumors recognition of antigenic determinants (TCR signaling, BCR (Supplementary Fig. 4b, c). Conversely, AT3 tumors treated with signaling, co-receptor signaling), as well as NK cell, interleukin NAM exhibited increased levels of CD8+ and CD4+ T cells, as and chemokine signaling (Fig. 6e). However, when we compared well as a higher CD8+/CD4+ T cell ratio as compared to the immunological infiltrate of untreated M/D-driven tumors at untreated tumors (Fig. 6f). Moreover, NAM-treated AT3 tumors 0.7–1cm2 surface area with that of their NAM-treated contained increased amounts of CD8+ cells (co)-expressing the

8 NATURE COMMUNICATIONS | (2020) 11:3819 | https://doi.org/10.1038/s41467-020-17644-0 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-17644-0 ARTICLE

a Control (n = 84) Prevention (n = 46) Treatment (n = 23) b Control (n = 13) NAM (n = 15) 100 2.5 100 p<0.0001 HR: 0.28 [0.11;0.71] p<0.0001 80 2.0 80 HR p value vs. control ) 2 60 0.75 [0.53; 1.06] p = 0.0946 1.5 60 0.54 [0.36; 0.83] p = 0.0046 40 1.0 40 OS (%) TTD (%) HR p value vs prevention Aeea (cm 20 0.55 [0.34; 0.89] p = 0.0110 0.5 20 0 0.0 0 0 50 100 150 0 10 20 30 0 20 40 60 80 100 Days after detection Days after treatment Days after treatment

c +2 d Ctrl DMBA AT3 Up: 2193 Down: 19 Up: 301 Down: 1244 Up: 26 Down: 393 NAM

–2

GO pathway p value Gene FC p value Gene FC p value Membrane 3.03E–15 Klrb1c 4.53 1.15E–04 Cd2 2.58 1.78E–01 Cd300lg ltga4 Nlrc3 4.24 2.79E–03 Akna 2.49 4.20E–02 lmmune system process 1.20E–04 Cdh2 Limch1 Pla2g2d 3.93 8.26E–03 Cd160 2.37 4.23E–02 Dock10 N4bp2 Positive regulation of TLR4 signaling pathway 5.20E–04 H2-Eb2 3.78 4.92E–03 Oas1b 2.32 1.57E–02 Dock2 Phka1 Tmem86b 3.71 3.12E–03 Slamf7 2.29 2.33E–02 lnterleukin-1 beta secretion 2.54E–03 Lax1 3.36 5.56E–03 ll10ra 2.24 3.34E–02 E130308A19Rik Pik3r4 Positive regulation of interleukin-8 secretion 6.88E–03 Ptpn22 3.33 5.55E–03 Cd48 2.23 1.62E–02 Fbxo21 Pik3r6 Map4k2 3.10 2.08E–03 Ly9 2.20 3.28E–02 Fmnl1 Plcb4 Integrin-mediated signaling pathway 8.03E–03 Fcna 3.09 1.19E–03 Ada 2.18 2.16E–02 Gimap8 Plpp3 Negative regulation of T cell activation 2.40E–02 Madcam1 3.03 3.67E–03 Lck 2.17 4.40E–02 Gm40813 Ppp1r12b Satb1 3.00 8.24E–03 Cd300lg 2.15 1.61E–02 Integral component of membrane 3.49E–02 Grb10 Prkcb Hmha1 2.98 1.08E–02 Klrd1 2.13 7.85E–03 Hdac9 Slc9a7 T cell activation 3.71E–02 Cd69 2.96 1.04E–02 Nlrc4 2.06 3.16E–02 Cd22 2.94 1.78E–02 Tlr6 1.77 3.83E–02 ll10ra//Mir7086 Syne3 Positive regulation of leukocyte migration 4.74E–02 Klra7 2.78 1.19E–02 Abcb1b 1.76 3.83E–02 Inpp5d Ttc7b Positive regulation of T cell differentiation 4.74E–02 Blnk 2.71 3.16E–02 Sfxn4 1.74 2.63E–02 e Molecular mechanisms of cancer Integrin signaling Hereditary breast cancer singaling lL-12 signaling and production in macrophages Estrogen dependent breast cancer signaling p53 signaling NK cell signaling T cell exhaustion signaling pathway Chemokine signaling B cell receptor signaling CD28 signaling in Th cells Nur77 signaling in T lymphocytes CXCR4 signaling Th1 and Th2 activation pathway Phosphatidylcoline biosisntesis lL-3 signaling T cell receptor signaling Choline biosisntesis lll lL-8 signaling lL-9 signaling Triacilglycerol biosynthesis PPARa/RXRa activation lL-6 signaling Fcγ receptor-mediated phagocytosis in macrophages and monocytes

p = 0.0063 f 15 p = 0.0304 8 p = 0.0001 8 5 5

/ n = 5 / + n = 7 n = 6 4 + 4 p = 0.0444 cells 6 cells 4 6 4 n = 9 n = 7

10 CD8 cells CD8 + 3 3 3 /10 /10 + + +

4 4 p = 0.0211 cells cells 4 4 /10 lL2 +

TNF 2 2 + +

n = 7 γ 10 10 CD8

5 CD8 γ + 2 n = 9 + 2 γ n = 9

CD8 1 1

n = 9 n = 9 lFN lFN TNF 0 lFN 0 0 0 0 Control NAM Control NAM Control NAM Control NAM Control NAM p<0.0001 p = 0.0278 15 2.0 15 / 4 p = 0.0086 2.0 n = 7 + cells p<0.0001 n = 7 p = 0.0421 4 n = 7 cells

1.5 n = 6 CD4 3 1.5 3 /10 n = 7 + cells

10 + 10 n = 9 + cells 3 /10 IL2 + +

1.0 cells 2 1.0 4 /10 CD4 + + n = 9 /CD4 +

5 5 10 CD4 TNF +

0.5 + 1 0.5

n = 9 γ TNF CD4 n = 9 n = 9 + CD8 γ TNF 0 0.0 0 lFN 0 0.0

Control NAM Control NAM lFN Control NAM Control NAM Control NAM

Fig. 6 NAM mediates immunotherapeutic effects on established BCs. a Time-to-death (TTD) of C57BL/6 mice subjected to M/D-driven carcinogenesis in control conditions or in the context of NAM supplementation with the drinking water, either from MPA pellet implantation (prevention) or tumor detection (treatment). Number of mice, hazard ratio (HR) and p values (two-sided log-rank) are reported. b Tumor growth and overall survival (OS) in C57BL/6 mice bearing established M/D-driven tumors that were maintained in control conditions or subjected to NAM supplementation with the drinking water. Tumor growth results are means ± SEM. Number of mice, HR and p values (two-way ANOVA corrected for row (time) and column (treatment) factors for tumor growth and two-sided log-rank for OS) are reported. c Non-supervised hierarchical clustering of genes differentially expressed in untreated (n = 4) vs NAM-treated (n = 4) M/D-driven tumors. Top upregulated genes, fold change (FC) and adjusted, two-sided p values are reported. Gene Ontology analysis and adjusted, two-sided p values for enrichment are indicated. d Comparison of private vs shared transcriptional changes induced by NAM treatment in M/D-driven tumors vs AT3 tumors established in C57BL/6 mice. The list of genes upregulated by NAM in both models is provided. See also Supplementary Fig. 3g. e Ingenuity Pathway Analysis of genes upregulated by NAM in both M/D-driven tumors AT3 tumors established in C57BL/6 mice. f Immune infiltration of AT3 tumors established in C57BL/6 mice that were maintained in control conditions or received NAM supplementation with the drinking water (starting at tumor detection) for 10 days. Results are means ± SEM plus individual data points. Number of mice and p values (one way-ANOVA plus Fisher LSD) are reported.

NATURE COMMUNICATIONS | (2020) 11:3819 | https://doi.org/10.1038/s41467-020-17644-0 | www.nature.com/naturecommunications 9 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-17644-0

a NAM (n = 11) b NAM (n = 11) α-CD4 + α-CD8 (n = 11) p = 0.1186, HR: 1.88 [0.78; 4.50] α-NK1.1 (n = 7) p = 0.0139, HR: 2.86 [0.86; 9.49] α-CD4 + α-CD8 + NAM (n = 12) p = 0.0010, HR: 3.10 [1.22; 7.88] α-NK1.1 + NAM (n = 9) p = 0.0608, HR: 2.17 [0.81; 5.81] 2.5 100 2.5 100 ) 2.0 ) 2.0 2 75 2 75 1.5 1.5 50 50 OS (%) 1.0 OS (%) 1.0 Area (cm Area (cm 0.5 25 0.5 25 p<0.0001 p<0.0001 0.0 0 0.0 0 0 10 20 30 0 20406080100 01020 30 0 20 40 60 80 100 Days after treatment Days after treatment Days after treatment Days after treatment

c NAM (n = 11) d α-lFNγ (n = 11) p = 0.6187, HR: 1.21 [0.52; 2.80] Control (n = 18) α-PD1 (n = 9) p = 0.1208, HR: 0.58 [0.27; 1.23] α-lFNγ + NAM (n = 16) p = 0.1623, HR: 1.63 [0.76; 3.46] NAM (n = 11) α-PD1 + NAM (n = 6) p = 0.0032, HR: 0.34 [0.15; 0.77] 2.5 100 2.5 100 p = 0.4324 ) ) 2.0 2.0 2 2 75 75 1.5 1.5 50 50 OS (%) 1.0 1.0 OS (%) Area (cm Area (cm 25 0.5 0.5 25 p<0.0001 p<0.0001 0.0 0 0.0 0 01020 30 04080 120 01020 30 0 20 40 60 80 100 Days after treatment Days after treatment Days after treatment Days after treatment

e p = 0.0959, HR: 0.61 [0.30; 1.25] Control (n = 21) α-PD1 (n = 10) p<0.0001, HR: 0.32 [0.16; 0.67] NAM (n = 27) α-PD1 + NAM (n = 18) p<0.0001, HR: 0.25 [0.11; 0.53] 2.0 100 )

2 1.5 75

1.0 50 OS (%)

Area (cm 0.5 25

0.0 0 0102030 20 30 40 50 Days after treatment Days after treatment

Fig. 7 The therapeutic effects of NAM depend on T cells and immune signaling. a–c Tumor growth and overall survival (OS) in C57BL/6 mice bearing established M/D-driven tumors that were maintained in control conditions (a–c) or subjected to NAM supplementation with the drinking water alone (a–c), or in the context of CD4+ and CD8+ T cell co-depletion (a), NK cell depletion (b), or IFNG neutralization (c). Tumor growth results are means ± SEM. Number of mice, hazard ratio (HR) and p values (two-way ANOVA corrected for row (time) and column (treatment) factors for tumor growth and two-sided log-rank for OS). Please note that NAM-treated tumors (red curves) in (a–c) are from the same experiment, also depicted in (d). d, e Tumor growth and OS in C57BL/6 (d) or BALB/c (e) mice bearing established M/D-driven (d) or TSA (e) tumors that were maintained in control conditions or subjected to NAM supplementation with the drinking water or systemic PD-1 blockage, alone or in combination. Tumor growth results are means ± SEM. Number of mice, hazard ratio (HR) and p values (two-way ANOVA corrected for row (time) and column (treatment) factors for tumor growth and two- sided log-rank for OS). Please note that NAM-treated tumors (red curves) in (d) are from the same experiment depicted in (a–c). effector molecules IFNG and tumor necrosis factor (TNF), and M/D-driven carcinomas resemble human HR+ BCs also in co-expressing IFNG plus the T cell mitogen interleukin 2 (IL2), as their limited sensitivity to PD-1 blockers8 (Fig. 7d). In this setting, compared to their untreated counterparts (Fig. 6f). Similarly, in the therapeutic activity of NAM was superior to that of PD-1 the AT3 model, NAM treatment was associated with increased blockers administered alone, and addition of PD-1 blockers to tumor infiltration by CD4+ T cells expressing TNF alone, TNF NAM treatment failed to ameliorate the therapeutic activity of the and IFNG, as well as TNF, IFNG, and IL2 (Fig. 6f). latter (Fig. 7d). In contrast, NAM synergized with PD-1 blockade in the TSA model (Fig. 7e). Similarly, NAM could be favorably combined with MTX-based chemotherapy in both the M/D- NAM mediates direct and indirect immunostimulatory effects. driven and AT3 models, and extended the survival of mice more To circumvent the heterogeneity of the M/D-driven model and than did either NAM or MTX alone (Supplementary Figs. 3B and acquire reliable mechanistic insights into the immunological 4H). Finally, we performed single-cell RNAseq (scRNAseq) on circuitries that underlie the therapeutic activity of NAM, we + performed depletion/neutralization experiments. This approach CD45 leukocytes infiltrating untreated vs NAM-treated TSA confirmed that the therapeutic activity of NAM depends on the tumors, revealing mild numerical alterations in the immune immune system, as demonstrated by experiments involving the infiltrate of NAM-treated tumors, including an increased co-depletion of CD4+ and CD8+ T cells, lymphocyte subsets frequency of T cells and monocytes (including DCs) coupled to expressing NK1.1 (NK cells, NKT cells and a subset of activated decreased abundance of tumor-associated macrophages (TAMs) CD8+ T cells)31, or IFNG neutralization (Fig. 7a–c). Similarly, the and NK cells (Fig. 8a). More importantly, NAM provoked ability of NAM to counteract the growth of AT3 tumors was considerable shifts in the transcriptional profile of all these four abolished by the co-depletion of CD4+ and CD8+ T cells as well immune cell populations (differentially expressed genes: 1884 in as by the neutralization of IFNAR1, IFNG, and IL17 (Supple- macrophages, 1541 in monocytes, 590 in T cells, and 283 in mentary Fig. 4d–g). NK cells), including multiple alterations supporting innate and

10 NATURE COMMUNICATIONS | (2020) 11:3819 | https://doi.org/10.1038/s41467-020-17644-0 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-17644-0 ARTICLE

a Macrophages (1288) T cells (207) Macrophages (1214) T cells (251) ControlMonocytes (1345) NK cells (82) NAM Monocytes (1758) NK cells (53)

–25 –25

0 0 tSNE_1 tSNE_1

25 25

50 50

–50 –25 0 25 –50 –25 0 25 tSNE_2 tSNE_2

b Macro p<2.2E–16 Mono p<2.2E–16 Macro p = 2.4E–14 Mono p<2.2E–16 Macro p = 9.6E–11 Mono p<2.2E–16 6 6 4 5 3 4

3 4 3 4 4 2 3 UMI (log) UMI (log) UMI (log) UMI (log) 2 2

UMI (log) UMI (log) 2 2 2 1 1 1 1 H2-D1 H2-D1 H2-DMb1 H2-DMb2 H2-DMb2 0 0 H2-DMb1 0 0 0 0 Control NAM Control NAM Control NAM Control NAM Control NAM Control NAM

Macro p = 2.1E–11 Mono p = 2.6E–10 Macro p = 8.9E–09 Macro p = 4.4E–03 Macro p = 7.2E–12 Macro p = 1.0E–10 6 6 6 5 5 5 4 4 4 4 4 4 3 3 3 UMI (log) UMI (log) UMI (log) UMI (log)

2 UMI (log) 2 UMI (log) 2 2 2 2 1 1 1 Cxcl9 Cxcl9 H2-Aa Cxcl10 H2-Eb1 0 H2-Eb1 0 0 0 0 0 Control NAM Control NAM Control NAM Control NAM Control NAM Control NAM

Macro p = 9.2E–11 Mono p = 1.1E–13 Mono p = 1.2E–06 Macro p = 6.4E–02 Macro p<2.2E–16 Mono p = 1.3E–12 3 4 6 5 4 4

3 4 3 3 2 4 3 2 2 2

UMI (log) 2 UMI (log) UMI (log) UMI (log) UMI (log) UMI (log) 1 2 1 1 1 1 lrf1 lrf1 ll10 ll10 ll1rn lfitm1 0 0 0 0 0 0 Control NAM Control NAM Control NAM Control NAM Control NAM Control NAM T cell p = 5.5E–05 T cell p = 1.8E–03 T cell p = 1.6E–02 T cell p = 2.8E–03 T cell p = 2.0E–02 T cell p = 1.2E–03 4 4 5 4 4 3

3 3 4 3 3 2 3 2 2 2 2 UMI (log) UMI (log) UMI (log) UMI (log) UMI (log) 2 UMI (log) 1 1 1 1 1 1 Cd3e Cd3e Cd3g Cd8a Faslg 0 0 0 0 Cd8b1 0 0 Control NAM Control NAM Control NAM Control NAM Control NAM Control NAM

NK p = 3.6E–02 T cell p = 9.5E–04 T cell p = 2.4E–02 T cell p = 4.6E–04 T cell p = 3.7E–02 T cell p = 4.6E–09 6 5 4 4 5 3

4 3 3 4 4 2 3 3 2 2 UMI (log) UMI (log) UMI (log) UMI (log) UMI (log) 2 UMI (log) 2 2 1 1 1 1 1 Lag3 Klrc1 Gzmb Gzmb Pdcd1 0 0 0 0 0 Cdkn1b 0 Control NAM Control NAM Control NAM Control NAM Control NAM Control NAM

cdp = 0.0260 p = 0.0037 p = 0.0106 4 n = 3 + 3 n = 4 4 8 p = 0.0361 n = 4 p<0.0001 p = 0.0091 +

n = 4 + n = 6 +

p = 0.0595 + n = 4 p = 0.4698 p = 0.0120 p = 0.2981 3

3 6 + CD45 n = 4 n = 3 + 2 n = 4 n = 6 p = 0.8346 p = 0.4701 p = 0.9371 CD45 p = 0.9976 p = 0.0958 2 n = 4 + 2 4 n = 3 n = 6 GZMB GZMB n = 4 n = 3 n = 4 +

+ p = 0.3228 on CD56

γ n = 3 γ 1 n = 4 + n = 3 γ

on CD56 1 on CD8 1 2 n = 6 lFN + lFN γ lFN on CD8 (%, fold change) (%, fold change) (%, fold (%, fold change) (%, fold 0 change) (%, fold 0 0 0 lFN NAM (mM) 0 0.1 1 2 3 NAM (mM) 0 0.1 1 2 3 NAM (mM) 0 0.1 1 2 3 NAM (mM) 0 0.1 1 2 3

e p = 0.0029 12 n = 2 20 n = 10 p = 0.0308 15 n = 3 p = 0.0078 8 n = 3 10 p = 0.4364 Ccl2 lfnb1 n = 10 4 n = 2 5 n = 10 n = 2 n = 2 n = 2 (fold change) (fold (fold change) (fold 0 0 NAM (mM) 0 0.1 1 2 3 NAM (mM) 0 0.1 1 2 3 adaptive immunity (Supplementary Fig. 5a–d and Supplementary CXCL10, and (3) proteins responding to type I IFN signaling, Data 1). In particular, monocytes and TAMs expressed increased such as interferon regulatory factor 1 (IRF1) and interferon levels of (1) multiple MHC molecules including H2-Aa, H2- induced transmembrane protein 1 (IFITM1) (Fig. 8b and DMb1, H2-DMb2, H2-D1, and H2-Eb1; (2) T-cell recruiting Supplementary Data 1). Conversely, monocytes and TAMs cytokines such as C-X-C motif chemokine ligand 9 (CXCL9) and infiltrating NAM-treated tumors exhibited lower expression of

NATURE COMMUNICATIONS | (2020) 11:3819 | https://doi.org/10.1038/s41467-020-17644-0 | www.nature.com/naturecommunications 11 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-17644-0

Fig. 8 NAM-treated tumors exhibit improved antigen presentation and superior cytotoxic functions. a tSNE plots of untreated and NAM-treated TSA tumors. Number of cells in each of the main four populations is reported. b Differential expression of genes involved in immune regulation in CD45+ cells isolated from untreated vs NAM-treated TSA tumors. Results are mean ± SEM plus individual data points. Number of cells independently analyzed for each gene is reported in (a), p values (two-sided Wilcoxon test) are indicated. See also Supplementary Data 1. c, d Percentage of IFNG+ and IFNG+GZMB+ amongst CD8+ T cells (c) and CD56+ NK cells (d) from peripheral blood mononuclear cells (PBMCs) of healthy donors subjected to non-specific activation overnight in the presence of the indicated concentrations of NAM. Results are means ± SEM plus individual data points. Number of biologically independent samples and p values (one way-ANOVA plus Fisher LSD, as compared to untreated cells) are reported. e Relative expression levels of Ifnb1 and Ccl2 in TSA cells cultured in control conditions or exposed to the indicated concentrations of NAM for 48 h. Results are means ± SEM plus individual data points. Number of biologically independent samples and p values (one way-ANOVA plus Fisher LSD, as compared to untreated cells) are reported.

immunosuppressive factors like IL10 and interleukin 1 receptor proliferative HER2−ER+VIM−) phenotype that fails to convert antagonist (IL1RN) (Fig. 8b and Supplementary Data 1). into basal (ER−VIM+) over time or upon treatment. Moreover, Amongst other changes, T cells infiltrating NAM-treated tumors M/D-driven tumors developing in immunocompetent mice exhibited (1) increased expression of genes involved in antigen resemble human luminal B BCs as they exhibit common PI3KCA recognition, including Cd3d, Cd3e, Cd3g, Cd8a and Cd8b1; (2) mutations32, poor immune infiltration coupled to limited sensi- upregulation of effector molecules including granzyme B (GZMB) tivity to ICBs8, and can be accelerated by nutritional interventions and Fas ligand (FASLG); (3) increased levels of surface activation that have been epidemiologically linked to increased propensity markers like PD-1, killer cell lectin-like receptor subfamily C, for HR+ BC in humans, such as a HFD17. One of the caveats of member 1 (KLRC1, best known as NKG2A), and lymphocyte this model relates to the fact that while ER+ BC is the most activating 3 (LAG3), and (4) reduced levels of negative cell cycle common form of BC in women, only a fraction of cases of ER+ regulators, such as cyclin dependent kinase inhibitor 1B BC is linked to the use of progestins11. Moreover, the M/D-driven (CDKN1B) (Fig. 8b and Supplementary Data 1). Despite their model develops in a rather heterogenous manner, which calls for limited number, NK cells infiltrating NAM-treated tumors also the use large experimental groups in support of statistical power, exhibited higher levels of GZMB as compared to NK cells from and displays a luminal B phenotype, which is common to other untreated tumors (Fig. 8b and Supplementary Data 1), which is mouse BC models including TSA and AT3 cells28,29, as well as indicative of (at least some degree of) functional activation. In line mice expressing MMTV-PyMT22. That said, heterogeneity is also with these transcriptional changes, Gene Ontology (Supplemen- a feature of human ER+ BC and to the best of our knowledge, no tary Fig. 6b) and Hallmarks (Supplementary Fig. 6b) analysis other model of HR+ BC enabling to investigate oncogenesis, revealed an enrichment of multiple gene sets linked to immune tumor progression and response to treatment in immunologically activation in the immune infiltrate of NAM-treated vs untreated intact hosts has been documented9. Thus, M/D-driven tumors TSA tumors, taken as a whole (CD45+ cells) as well as in its stand out as a privileged model for the study of BC immuno- specific components (macrophage, monocytes, T cells, NK cells). surveillance, resistance to (immuno)therapy, and sensitivity to Moreover, immunofluorescence microscopy and RT-PCR corro- nutritional interventions that operate by immunological borated the ability of NAM to increase the abundance of immune mechanisms. effector cells (CD8+ T cells) and transcripts (Gzmb, Ifng and Prf1, Interestingly, M/D-driven oncogenesis appears to be largely although the latter sub-significantly) in the microenvironment of controlled by an immunological network involving interferon TSA and M/D-driven tumors (Supplementary Fig. 7a, b). signaling and NK cell functions. This is at odds with other models Altogether, these findings lend additional support to the notion of endogenous-driven carcinogenesis, such as methylcholan- that NAM mediates therapeutic effects that depend on the threne (MCA)-dependent fibrosarcomas33 as well as with immune system. numerous instances of spontaneous carcinogenesis that are To obtain further mechanistic insights into this possibility, we accelerated by the lack of T cells or their molecular effectors34, exposed peripheral blood mononuclear cells (PBMCs) from potentially constituting a peculiarity of HR+ BCs. In this setting, healthy donors to phorbol 12-myristate 13-acetate (PMA) plus NAM supplementation with the drinking water not only delays ionomycin as non-specific activation stimuli, alone or in the the manifestation of M/D-driven tumors when provided as a presence of NAM, finding that the latter favors the accumulation prophylactic intervention, but also mediates therapeutic effects of both T cells and NK cells expressing IFNG alone or co- when initiated at tumor detection. In both these settings, the expressing IFNG and GZMB (Fig. 8c, d). We excluded the effects of NAM depend on CD8+ T lymphocytes but less so on possibility that NAM could mediate robust cytotoxic effects on NK cells. These findings raise the interesting possibilities that M/ malignant cells, even at highly supraphysiological concentrations D-driven carcinogenesis in mice (and perhaps HR+ mammary (Supplementary Fig. 7c), but identified the unsuspected ability of carcinogenesis in women) may be under natural immuno- NAM to drive transcription from the Ifnb1 locus, as well as from surveillance by NK cells35, and that NAM can be harnessed to re- the Ccl2 locus (Fig. 8e), which encodes a potent chemoattractant enable T cell-dependent tumor control in both prophylactic and for monocytes. Altogether, these findings are in line with therapeutic settings. Thus, established HR+ tumors may exhibit a scRNAseq and efficacy data, and suggest that NAM mediates dysfunctional NK cell compartment potentially sensitive to both direct and indirect immunostimulatory effects that impinge reactivation by recently developed antibodies specific for on T cell activation. NKG2A36. Further supporting this notion, while MCA-driven fibrosarcomas developing in immunocompetent animals are Discussion generally transplantable into both immunocompetent and Here, we demonstrate that M/D-driven tumors recapitulate key immunodeficient hosts33, the same does not apply to M/D-driven immunobiological features of human HR+ BCs, which account tumors, which are usually rejected by immunocompetent animals, for the majority of BC1. M/D-driven tumors emerge in immu- irrespective of the immunological competence of the host in nocompetent hosts via an ER-dependent oncogenic process, and which they originally developed. This suggests that evolving M/ progress as they evade immunosurveillance, culminating with the D-driven tumors are poorly edited by T lymphocytes even in establishment of mammary tumors with a luminal B (highly immunocompetent hosts, and hence evolve by accumulating

12 NATURE COMMUNICATIONS | (2020) 11:3819 | https://doi.org/10.1038/s41467-020-17644-0 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-17644-0 ARTICLE antigenic determinants that prevent engraftment in tumor-naïve Finally, NAM might also act as an inhibitor of immunosup- immunocompetent mice. The precise molecular mechanisms pressive NAD+-consuming enzymes that generate NAM as a underlying the inability of T cells to control M/D-driven carci- product of catalysis, such as CD38 (ref. 52) or poly(ADP-ribose) nogenesis and edit evolving M/D-driven tumors remain to be polymerase 1 (PARP1). Intriguingly, PARP1 has recently been elucidated. shown to limit type I IFN secretion by cancer cells with defects in In the course of this study, we discovered that NAM is parti- the DNA damage response53,54, and our findings indicate that cularly efficient at controlling the growth of various mouse NAM not only favors type I IFN secretion by TSA cells main- models of luminal B BC, including M/D-driven tumors as well as tained in vitro, but also (1) elicits signatures of type I IFN sig- TSA and AT3 tumors established in immunocompetent hosts, naling in immune cells infiltrating TSA tumors established in both as a standalone agent and in combination with clinically immunocompetent hosts, and (2) mediates prophylactic and approved agents (e.g., anthracyclines and PD-1 blockers). The therapeutic effects that depend (at least in part) on type I IFN dose of NAM employed in this study (0.5% w/w with the drinking responses. That said, whether PARP1 inhibition underlies the water) admittedly results in supraphysiological levels of blood- beneficial effects of NAM against mammary tumors remains to be borne NAM (16.3 µg mL−1). Although peak NAM concentrations experimentally verified. of ~2.3 µg mL−1 have been documented in healthy volunteers Irrespective of these hitherto untested possibilities, NAM receiving 2 g nicotinic acid (a precursor of NAM) per os20, supplementation stands out as a safe and effective option for the whether higher steady-state NAM levels can be safely achieved in prevention and treatment of HR+ BC. Prospective, randomized humans remains unexplored. That said, NAM has been suc- clinical trials clarifying the preventive are therapeutic activity of cessfully used in in randomized clinical trials to boost the efficacy NAM are urgently awaited. of radiation therapy in patients with bladder and laryngeal carcinoma37,38, as well as for the prevention of actinic keratosis Methods and non-melanoma skin cancers39. The latter effect has been Cell lines. Mouse mammary adenocarcinoma TSA cells (#SCC177), mouse linked with reduced tumor infiltration by CD68+ inflammatory mammary carcinoma AT3 cells (#SCC178) and mouse fibrosarcoma MCA205 cells macrophages40, which is in line with our observations in the (#SCC173) were obtained from Millipore Sigma. All cell lines were cultured at 37 °C under 5% of CO2, in the appropriate medium containing 10% fetal bovine therapeutic setting. In a preclinical model of benzo(a)pyrene- serum (FBS), 100 U mL−1 penicillin sodium and 100 µg mL−1 streptomycin sulfate induced lung carcinogenesis, oral (but not nasal) administration (both included in #15070063 from Thermo Fisher). TSA cells were cultured in of NAM also demonstrated oncopreventive effects41. Thus, NAM Dulbecco’s Modification of Eagle’s Medium (DMEM, #11960044 from Thermo might exert a relative broad oncosuppressive action, contrasting Fisher), AT3 cells in RPMI 1640 medium (#11875119, Thermo Fisher), and MCA205 cells in RPMI 1640 medium supplemented as above plus 1 mM sodium with other vitamins such as pyridoxine and calcitriol, which failed pyruvate (#11360070, Thermo Fisher) and 1 mM HEPES buffer (#15630106, to enable oncoprevention in observational or interventional Thermo Fisher). All cell lines were routinely checked for Mycoplasma spp. con- clinical studies42–44. Of note, haploidentical or mismatched tamination by the PCR-based LookOut® Mycoplasma PCR Detection Kit related donor NK cells expanded in the presence of NAM are (#MP0035, Millipore Sigma). currently being investigated for the treatment of refractory/ relapsed multiple myeloma and non-Hodgkin’s lymphoma Primary cell cultures. After sacrifice, cells from M/D-driven tumors developed in WT or Rag2−/− mice were surgically recovered and washed in DMEM/F12 (NCT03019666), corroborating the elevated potential of NAM as medium (#10565042, Thermo Fisher) supplemented with 100 U mL−1 penicillin an immunostimulant for clinical applications. sodium, 100 µg mL−1 streptomycin sulfate and 100 µg mL−1 gentamycin. They As for mode of action, it appears plausible that NAM acts as were then dissociated in wash medium supplemented with 0.15% type A col- a therapeutic agent by re-establishing T cell-dependent lagenase (#10103578001, Millipore Sigma) for 30 min at 37 °C (with occasional mechanical dispersion). After dissociation, cells were cultured in flasks pre-coated immunosurveillance via both direct (T cell activation) and with 0.1% gelatin (4 h, 37 °C), in DMEM/F12 medium supplemented with 2% FBS, indirect (type I IFN and CCL2 secretion by cancer cells) 50 µg mL−1 gentamycin sulfate (#15750078, Thermo Fisher), 10 ng mL−1 insulin mechanisms. However, the precise molecular cascade(s) (#I0516, Millipore Sigma) and 5 ng mL−1 epithelial growth factor (EGF; #SRP3196 underlying such beneficial effects of NAM remain(s) to be from Thermo Fisher). elucidated. Hydroxycarboxylic acid receptor 2 (HCAR2), a G Mouse mammary gland organoids (MMGOs). For MMGO handling, plasticware protein-coupled receptor for nicotinic acid, has been reported −1 to suppress mammary oncogenesis in mice by inhibiting cancer was pre-coated with PBS supplemented with 2.5 mg mL bovine serum albumin 45 fi (BSA, #700-100P from Gemini Bio-Products). Mammary glands were collected, cell survival . However, HCAR2 has a 1000-fold lower af nity and lymph nodes removed. Mammary glands were then minced with scissors and for NAM than for nicotinic acid46, and although some degree of digested in DMEM/F12 medium supplemented with 5% FBS, 50 µg mL−1 genta- − NAM conversion into nicotinic acid by the gut microbiota micin sulfate, 5 ng mL 1 insulin, 0.04% (w/v) Trypsin-EDTA (#15400054, Thermo −1 cannot be excluded47, hepatic metabolism efficiently removes Fisher) and 2 mg mL collagenase A (#10103586001, Roche) for 1 h at 37 °C at 47 100 rpm in a HulaMixer sample mixer (Life Technologies). Thereafter, mammary nicotinic acid from the circulation by forming NAM .More- glands were subjected to vigorous shaking to disrupt adipocytes, and cells were over, mouse BC cells exposed to highly supraphysiological pelleted at 520 g for 10 min. After further washing in DMEM/F12 medium, both NAM concentrations up to 80 mM did not undergo cell death the cell pellet and the fat layer were recovered. The latter was mechanically dis- considerably, ruling out cytotoxicity from the mechanisms sociated, pelleted again, and then added to the cell pellet. Thereafter, cells were resuspended for 5 min in DMEM/F12 medium supplemented with 4 U mL−1 through which NAM mediates therapeutic effects in our models DNase I (#DPRF, Worthington Biochemical), pelleted, and cleared by differential (further corroborated by the inability of NAM to slow down centrifugation. Mammary organoids were finally collected and embedded in tumor growth in vivo upon T cell depletion). Matrigel® Matrix (#356255, Corning) for 3D culture. These observations not only cast doubts on the actual impli- cation of HCAR2 in the beneficial effects of NAM against Mice.WTorgeneticallymodified C57BL/6 and BALB/c female mice (Mus musculus) mammary tumors, but also argue against the involvement of of 6-15 weeks of age were employed. Mice were maintained in standard specific pathogen-free (SPF) housing conditions (20 ± 2 °C, 50 ± 5% humidity, 12h-12h light- sirtuin 1 (SIRT1) inhibition, which has been linked to the death of dark cycles, food and water ad libitum), unless specified as per study design. Animal 48 human BC cells exposed to high-dose NAM . Of note, low NAM experiments followed the Federation of European Laboratory Animal Science Asso- concentrations have been reported to activate (rather than inhi- ciation (FELASA) guidelines, were in compliance with the EU Directive 63/2010 bit) SIRT1 in numerous settings49, and SIRT1 functions appear to (protocol 2012_034A) and were approved by institutional ethical committees for be required for robust immune responses in a variety of animal experimentation at Gustave Roussy (no. 2016031417225217), Centre de 50,51 Recherche des Cordeliers (no. 2016041518388910), and Weill Cornell Medical College settings . Additional work is required to assess the impact of (no. 2017-0007 and 2018-0002). WT C57BL/6 and BALB/c mice were obtained from SIRT1 on the immunosurveillance of luminal B BC. Harlan France or Taconic Farms, Rag2−/−Il2rg−/−, Rag2−/−,Casp1−/−, Il21−/−,

NATURE COMMUNICATIONS | (2020) 11:3819 | https://doi.org/10.1038/s41467-020-17644-0 | www.nature.com/naturecommunications 13 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-17644-0

Il21r−/−, Il1r1−/−, Nlrp3−/−, Ifng−/−,andIfnar1−/− mice were obtained from The Millipore Sigma), followed by rinsing with PBS (4×) and blockage of non-specific Jackson Laboratory, Il22−/− and Myd88−/− mice were obtained from CNRS UMR binding sites with 3% BSA in TBS-Tween for 30 min at RT. Slides were then 7355 (Orleans, France). MMTV-PyMT mice were obtained from Prof. M.F. Krummel incubated overnight at 4 °C with the following primary antibodies: anti-ESR1- (UCSF, San Francisco, CA, USA), Becn1+/− mice from Prof. B. Levine (UT South- AlexaFluor488 (1:20, #sc-8005 from Santa Cruz Biotechnology) and anti-VIM western Medical Center, Dallas, TX, USA), Atg5fl/fl KRT5-Cre mice Prof. J. Penninger (1:250, #GTX100619, from GeneTex). Next, slides were rinsed 3× in TBS-Tween, (Institute of Molecular Biotechnology, Vienna, Austria) and originally from Prof. incubated with an AlexaFluor594-conjugated goat anti-rabbit antibody (1:500; Noboru Mizushima (University of Tokyo, Tokyo, Japan), and both ERα AF-20 and #ab150084 from Abcam) for 30-60 min at RT, rinsed again 3× in TBS-Tween and ERC451A mice were kindly provided by Prof. F. Lenfant (INSERM U1048, Toulouse, 1× in deionized water, and incubated with Post-Detection Conditioner (part of France). In all experiments, mice were routinely monitored for tumor growth and #MB-L, MaxVision Biosciences Inc.) for 5 min at RT. Finally, slides were incubated euthanatized when tumor surface reached 200-250 mm2 (ethical endpoint), or in the in 5 µg mL−1 Hoechst 33258 (#H3569, Thermo Fisher) for 10 min at RT, mounted presence of overt signs of distress (e.g., hunching, anorexia, tumor ulceration). with ProLong™ Diamond Antifade Mountant (#P36961, Thermo Fisher), and imaged on an Eclipse TiE Motorized Digital Fluorescence Microscope operated by NIS-Elements AR v. 4.11 (Nikon). Oncogenesis. Fifty mg slow-release (90 days) MPA pellets (#NP-161, Innovative Research of America) were implanted subcutaneously by surgery into 6–9-weeks- old female mice (day 0). Mice were administered 200 µL of a 5 mg mL−1 7,12- Immunofluorescence microscopy—TSA tumors. Freshly excised TSA tumors dimethylbenz[a]anthracene (DMBA; #D3254, from Millipore Sigma) solution in were fixed in zinc fixative (#552658, BD Biosciences) prior to embedding in par- corn oil (#C8267, Millipore Sigma), by oral gavage once a week on weeks 1, 2, 3, 5, affin. After deparaffinization and dehydration, slides were boiled in de-cloacker 6, and 7 after implantation of the MPA pellet. media (#RD913M, Zytomed), treated with peroxide block (# ZUC019-008, Zytomed), normal goat serum (#S-1000-20, Vector Laboratories) and stained with primary antibodies specific for CD8A (1:2000, #14-0808-80, from Thermo Fisher) Transplantable breast cancers. For tumorigenicity assays, 1.0 × 106 AT3, 0.1 × and secondary ImmPRESSTM HRP anti-rat IgG antibodies (ready for use, #MP- 106 TSA, 0.3 × 106 MCA205 cells, or 0.5 × 106 cells established from M/D-driven − − 7444-15, from Vector Laboratories). Detection was performed with the Opal650 tumors evolving in Rag2 / or C57BL/6 mice (MGT cells) were injected sub- OPAL reagent (1:300, part of #NEL811001KT, Akoya Biosciences). After nuclear cutaneously or in the mammary fat pad of syngeneic mice. counterstaining with 4′,6-diamidino-2-phenylindole (DAPI, 2 drops mL−1, part of #NEL811001KT, from Akoya Biosciences), slides were mounted and evaluated on a Depletion/neutralization procedures. All antibodies blocking or neutralizing Vectra Polaris™ Automated Quantitative Pathology Imaging System operated by specific immune cell populations, cytokines or cytokine receptors and their cor- embedded software Vectra Polaris v.1 (Perkin Elmer). Whole slides were scanned responding isotype controls were acquired from BioXCell (West Lebanon): CD4 first, followed by acquisition of multiple regions of interest covering almost the (clone GK1.5, #BE0003-1), CD8α (clone 2.43, #BE0061), NKG2D (clone HMG2D, whole tumor surface with Phenochart v. 1.0.8, which were analyzed with inForm v. #BE0111), NK1.1 (clone PK136, #BE0036), PD-1 (clone RMP1-14, #BE0146), IFNγ 2.4.6 (Akoya Biosciences)55,56. (clone R4-6A2, #BE0054), IFNAR1 (clone MAR1-5A3, #BE0241), IL-17A (clone 17F3, #BE0173). Antibodies in the amount of 0.2-0.4 mg/mouse were administered Cell death. Cell death was quantified by flow cytometry on a MACSQuant analyzer intraperitoneally once every 8 days or, in the case of anti-PD1, every 3 days for 3 operated bv MACSQuantify™ v. 2.11 (Miltenyi) upon staining with 0.5 µg mL−1 times. For vaccinations experiments, ANXA1 neutralization was achieved with a propidium iodide (PI, #P4170 from Millipore Sigma) staining57. Gating procedure specific antibody (clone 29, #610066 from BD Biosciences). is exemplified in Supplementary Fig. 8.

Food and drugs. Tamoxifen (#T5648, Millipore Sigma) and NAM (#N0636, RT-PCR. Mouse Ccl2 and Ifnb1 levels were quantified relative to Rpl13 levels by Millipore Sigma) were administrered in the drinking water at concentrations of − 2-steps RT-PCR based on commercial primer sets from Bio-Rad (Ccl2, unique 10 μgmL 1 and 0.5% (w/v), respectively. For starvation experiments, mice under- assay ID qMmuCED0003785; Ifnb1, unique assay ID qMmuCED0002606; Rpl13a, went biweekly 24-h-long fasting cycles (with water ad libitum). Pellets for NR − unique assay ID qMmuCED0040629), the SuperScript™ VILO™ Master Mix supplementation (3.33 g Kg 1 pellets), HSD and HFD, and the corresponding (#11755500, from Thermo Fisher) and the iTaq Universal SYBR Green Supermix placebo pellets were obtained from Ssniff Spezialdiäten GmbH. Pyridoxine hydro- − − − (#1725121, Bio-Rad). Alternatively, mouse Gzmb, Ifnb1, Ifng and, Prf1 levels chloride (187 mg Kg 1), calcitriol (300 µg Kg 1), and spermidine (50 mg Kg 1) were quantified relative to Rpl13 levels by 1-step RT-PCR based on commercial (all from Millipore Sigma, cat. #P9755, #C0225000 and # S2626, respectively) were − primer and probe sets (Gzmb, #Mm00442837_m1; Ifnb1, #Mm00439552_s1; Ifng, delivered i.p. in 100 μL PBS once a week. For in vivo experiments, 5.17 mg Kg 1 #Mm01168134; Pfr1, #Mm00812512_m1; Rpl13, #Mm02526700_g1) and the MTX (#M2305000, Millipore Sigma) was administered i.p. in 200 µL PBS once at TaqMan™ Fast Virus 1-Step Master Mix (#4444434), all from Thermo Fisher. randomization. Solutions and bottles were changed three times a week. Amplifications were run on a 7500 RT-PCR system (Applied Biosystems) operated by embedded software v. 2.3 as per thermal protocols provided by the primer μ Δ ΔΔ Vaccination experiments. Cell lines or MMGOs were exposed in vitro to 1 M manufacturer. RT-PCR data was normalized according to the Ct or Ct MTX or 20 Gy irradiation and incubated for 24 h to obtain around 50–60% of methods. dying cells, followed by injection into the mammary fat pad of WT mice 1 week prior, in the 4th week of (break week), and 2 days after, DMBA gavage. ANXA1 CRISPR/Cas9. TSA cells were transfected with a control commercial CRISPR-cas9 blockage was achieved by injecting 12.5 µg of the blocking antibody (or the cor- plasmid (#CRISPR06-1EA, from Millipore Sigma) or with a CRISPR-cas9 plasmid respondent isotype control) in the mammary fat pad 24 h before and after vacci- − specific for Atg7 (custom-made by Millipore Sigma based on #CRISPR06-1EA), nation, and by complementing the vaccine with 250 µg mL 1 of the same antibody. using the TransIT-CRISPR® reagent (#T1706, Sigma Aldrich). GFP+ clones were sorted on an FACSAria II Sorter operated by FACSDiva™ v. 6.1.3 (from BD Histological studies. Tissues were fixed in 4% formaldehyde (#F8775, Millipore Biosciences) into 96-well plates, followed by clone selection and confirmation of Sigma) overnight and kept in ethanol 70% until inclusion in paraffin. Tissue ATG7 status by immunoblotting. processing was done by the Electron Microscopy & Histology Service of Weill Cornell Medicine. Fixed tissue was processed and embedded using a Tissue-Tek ShRNA transfection. AT3 cells were stably transfected with commercial plasmid VIP® 6 AI Vacuum Infiltration Processor and Tissue Embedding Console (Sakura). encoding an Atg7-specific shRNA (#TG504956, Origene Technologies) using the Processing consisted of multiple ethanol dehydration steps of 1.5 h each: 70, 85, 95, FuGen® HD transfection reagent (#E2311, Promega), as per manufacturer 95, 100, and 100%. This was followed by two exchanges of HistoChoice® Clearing instructions. ATG7 status was confirmed by immunoblotting. Agent (#H2779-1L, Millipore Sigma) and three exchanges of paraffin (2 h each). All steps were performed using vacuum and pressure. The paraffin steps were per- formed at 60 °C, all other steps were performed at room temperature (RT). The NAM quantification. Targeted LC/MS analyses of NAM were performed on a Q blocks were sectioned at 7-μm thick, collected onto positively charged glass slides, Exactive™ Orbitrap Mass Spectrometer (Thermo Fisher) coupled to a Vanquish™ and stained with Hematoxylin and Eosin Stain Kit (#H-3502, Vector Laboratories). HPLC system operated by Xcalibur v. 4.0.27.19 (Thermo Fisher). A SeQuant® The histology of M/D-driven tumors was compared with archival images of HR+ ZIC®-HILIC column (2.1 mm i.d. × 150 mm, Merck) was used for metabolite BC tissues from Weill Cornell Medicine (kindly provided by Dr. Syed Hoda). separation. Flow rate was set at 150 μL/min. Buffers consisted of 100% acetonitrile for mobile A, and 0.1% NH4OH/20 mM CH3COONH4 in water for mobile B. Gradient ran from 85% to 30% A in 20 min followed by a wash with 30% A and re- Immunofluorescence microscopy—M/D-driven tumors. Slides with M/D-driven equilibration at 85% A. NAM was identified in positive ion mode on the basis of tumors obtained for histological studies were deparaffinized and re-hydrated by exact mass within 5 ppm and standard retention time. Absolute quantitation was three incubations (5 min each) in xylenes (#214736, Millipore Sigma) followed by performed using a NAM-based standard curve. serial (2×, 3 min each) incubations in 100, 90, 80, and 70% ethanol. Auto- fluorescence was blocked with the MaxBlock™ Autofluorescence Reducing Reagent (#MB-L, MaxVision Biosciences Inc.) for 10 min at RT, followed by wash in 60% Immunoblotting. For detection of autophagy biomarkers in vivo, tissues were snap ethanol and incubation in deionized water (3×, 5 min each). Antigen retrieval was frozen, and then mechanically disrupted before cell lysis. For ATG7 detection, cell achieved by boiling sections for 50 min in 1X Citrate Buffer (pH 6, #C9999, from lines were subjected to standard lysis procedures. In both cases, 50 µg proteins were

14 NATURE COMMUNICATIONS | (2020) 11:3819 | https://doi.org/10.1038/s41467-020-17644-0 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-17644-0 ARTICLE separated on NuPAGE® Novex® Bis-Tris 4–12% pre-cast gels (Invitrogen) and To measure the populations of αβ T lymphocytes and their activation/ electrotransferred to polyvinyldifluoride (PVDF) membranes (Millipore Sigma). exhaustion status, cells were stained with LIVE/DEAD™ Fixable Yellow dye and Fc Membranes were blocked with 0.05% Tween 20 (v/v in TBS) supplemented with receptors were blocked as described above. Then, a cell surface staining was 5% non-fat powdered milk for 1 h and incubated overnight with primary antibody realized with the following fluorescent antibodies: anti-CD3-APC (1:200, clone specific for MAP1LC3B (1:1000, #2775 from Cell Signaling Technology), SQSTM1 17A2, #17-0032-82 from Thermo Fisher), anti-CD8-PE (1:400, clone 53-6.7, (1:1000, #5114 from Cell Signaling Technology), ATG7 (1:1000, clone D12B11, #553032 from BD Biosciences), anti-CD4-PerCP-Cy5.5 (1:400, clone RM4-5, #45- #8558 from Cell Signaling Technology; or 1:3000, clone ATG7-13, #SAB4200304 0042-82 from Thermo Fisher), anti-CD25 PE-Cy7 (1:100, clone PC61.5, #25-0251- from Millipore Sigma), or ACTB (1:1000, clone 13E5, #4970 from Cell Signaling 82 from Thermo Fisher), and anti-PD-1-APC/Fire750 (1:100, clone 29 F.1A12, Technology; or 1:2000, clone 8H10D10, #3700 from Cell Signaling Technology), at #135239 from BioLegend). Cells were then fixed and permeabilized in 1X Foxp3 4 °C. Primary antibodies were detected with horseradish peroxidase (HRP)-con- Transcription Factor Staining Buffer (part of #A25866A, from Thermo Fisher). jugated anti-mouse (#NA931, from GE Healthcare Life Sciences, 1:5000) or anti- Finally, an intranuclear staining was performed with anti-FOXP3-FITC (1:50, clone rabbit (#NA934, from GE Healthcare Life Sciences, 1:5000) secondary antibodies FJK-16s, #11-5773-82 from Thermo Fisher) and anti-Ki67-AlexaFluor700 (1:100, and revealed with the Pierce™ ECL Plus chemiluminescent substrate (#32132, clone B56, #561277 from BD Biosciences). Gating procedure is exemplified in Thermo Fisher) on a C600 Gel Doc & Western Imaging System operated by cSeries Supplementary Fig. 8. Capture v. 1.6.8.1110 (Azure Biosystems). Alternatively, cells were stained with LIVE/DEAD™ Fixable Yellow dye and Fc receptors were blocked as described above, then fixed in Cytofix/Cytoperm™ buffer after surface staining with anti-CD3-FITC (1:800, clone 17A2, #11-0032-82 from Flow cytometry—Tumor and mammary gland processing. M/D-driven tumors Thermo Fisher), anti-NK1.1-PerCP-Cy5.5 (1:100, clone PK136, #551114 from BD and mammary glands were recovered, cut with scissors and digested in RPMI Biosciences), anti-B220-V450 (1:100, clone RA3-6B2, # 560473 from BD medium supplemented with 26.67 µg mL−1 Liberase™ (#5401119001, Millipore − Biosciences) and anti-CD19-APC-Vio770 (1:50, clone REA749, #130-111-886 from Sigma) and 0.0167 MU mL 1 DNase I for 30 min at 37 °C. RPMI medium sup- Miltenyi). Gating procedure is exemplified in Supplementary Fig. 8. plemented with 10% FBS was then added to stop enzymes activity, and tumors were crushed on a 100 µm cell strainer with the back of a syringe plunger. After washing, cells were pelleted for 5 min at 300 g and resuspended in 10 mL RPMI RNASeq. Tissues of interest (including tumors) were recovered, stabilized in medium supplemented with 10% FBS. RNAlater RNA Stabilization Reagent (#R0901, Millipore Sigma) and RNA was extracted using RNeasy Mini kits (#74104, Qiagen), as per manufacturer’s instructions. Sequencing, data quality, reads repartition (e.g., for potential riboso- Flow cytometry—spleen and lymph node processing. Spleens and inguinal mal contamination) was performed by GenoSplice technology (www.genosplice. lymph nodes were collected and crushed between two microscope glass slides. Cells com) using Mouse Gene v. 1.0 ST Array for data presented in Fig. 1, or FastQC v. were resuspended in RPMI medium, filtered through 70 µM MACS® SmartStrai- 0.11.2, Picard-Tools v. 1.119, Samtools v. 1.0, and RSeQC v. 2.3.9 for data presented ners (#130-098-462, Miltenyi) and pelleted for 5 min at 300 g. Splenocytes and in Fig. 6 and Supplementary Fig. 3. lymph node cells were finally resuspended in 2 mL and 500 µL of RPMI supple- mented with 10% FBS, respectively. Single-cell RNAseq—cell isolation. TSA tumors were recovered from three mice per group, and dissociated using the Tumor Dissociation Kit, mouse (#130-096- Flow cytometry—in vitro T and NK cell phenotyping. PBMCs were isolated from 730, Miltenyi) and gentleMACS™ Octo Dissociator (Miltenyi), following manu- a 10 mL blood aliquot from healthy donors (following ethical guidelines of the facturer’s instructions. Upon pooling, samples were subjected to magnetic University Hospital Motol, Prague, and upon collection of informed consent separation of CD45+ cells using CD45 MicroBeads (#130-052-301, Miltenyi) and a forms) by Ficoll® Paque PLUS (#17-1440-02, GE Healthcare Life Sciences) gradient ’ − MACS Separator (Miltenyi), as per manufacturer s instructions. Viability was centrifugation. PBMCs were then cultured overnight in the presence of 50 ng mL 1 tested upon staining a cell aliquot with 0.05% trypan blue (#T6146, Millipore phorbol 12-myristate 13-acetate (PMA, # P8139 from Millipore Sigma) plus 1 μg − Sigma) in PBS on a Cellometer AutoT4 cell counter (Nexcelom Bioscience), and mL 1 ionomycin (#I3909, Millipore Sigma) alone or combined with NAM, fol- CD45+ cells were resuspended in PBS supplemented with 0.1% BSA. lowed by 3-h incubation with brefeldin A (1:1000, # 420601 from BioLegend). Cells were then washed in PBS, stained with either anti-CD45-PE (1:15, clone HI30, #MHCD4517 from Thermo Fisher) anti-CD3-AlexaFluor700 (1:20, clone MEM- Single-cell RNAseq—sample preparation and processing. Library preparation 57, #A7-202-T100 from EXBIO) and anti-CD8-HV500 (1:20, clone RPA-T8, for Single Cell 3′ RNA-seq v.2, sequencing and post-processing of the raw data was #560775 from BD Biosciences) monoclonal antibodies, or anti-CD45-V500 (1:15, performed at the Epigenomics Core at Weill Cornell Medicine, as follows. Single- clone HI30, #560777 from BD Biosciences), anti-CD3-AlexaFluor700 (1:20, clone cell RNAseq libraries were prepared according to 10X Genomics specifications OKT3, # A7-631-T100 from EXBIO), and anti-CD56-ECD (1:15, clone N901, (Single Cell 3′ Reagent Kits v.2 User Guide PN-120236, 10x Genomics). Two #B49214 from Beckman Coulter) monoclonal antibodies, then fixed in independent cell suspensions (~60% viable) at a concentration of ~900 cells µL−1 eBioscience™ fixation/permeabilization buffer (part of #88-8824-00, Thermo were loaded onto to the 10X Genomics Chromium platform to generate barcoded Fisher), further permeabilized with eBioscience™ permeabilization buffer (part of single-cell GEMs, targeting about 3000 single cells per sample. GEM reverse #88-8824-00, Thermo Fisher) and intracellularly stained with anti-IFNγ-PE-Cy7 transcription (53 °C for 45 min, 85 °C for 5 min; held at 4 °C) was performed in a (1:100, clone 4S.B3, #25-7319-82 from Thermo Fisher) and anti-GZMB-BV421 C1000 Touch Thermal cycler with 96-Deep Well Reaction Module (Bio-Rad). After (1:25, clone GB11, #563389 from BD Biosciences) monoclonal antibodies. Flow reverse transcription, GEMs were ruptured and single-stranded cDNA was cleaned cytometry was performed on a LSRFortessa™ Flow Cytometer operated by up with DynaBeads MyOne Silane Beads (#37002D, Thermo Fisher). cDNA was FACSDiva™ v. 6.2 (BD Biosciences), and data were analyzed with FlowJo v. 9.5.3 amplified for 12 cycles (98 °C for 3 min; 98 °C for 15 s, 67 °C for 20 s, 72 °C for (TreeStar, Inc.). Gating procedure is exemplified in Supplementary Fig. 8. 1 min) using the C1000 Touch Thermal cycler with 96-Deep Well Reaction Module. Quality of the cDNA was assessed using a Bioanalyzer 2100 (Agilent), obtaining a product of about 1693bp. This cDNA was enzymatically fragmented, Flow cytometry—phenotyping of the immune infiltrate. For immune cell phe- end-repaired, A-tailed, subjected to a double-sided size selection with SPRIselect notyping, 100 µL of spleen, lymph node, tumor, and mammary gland suspensions beads (#B23317, Beckman Coulter) and ligated to adaptors provided in the kit. underwent one of the following procedures, after which stained samples were A unique sample index for each library was introduced through 13 cycles of PCR acquired on an LSR II Flow Cytometer operated by FACSDiva™ v. 6.1.3 (BD amplification using the indexes provided in the kit (98 °C for 45 s; 98 °C for 20 s, Biosciences) and analyzed with FlowJo v. X.6.2. 54 °C for 30 s, and 72 °C for 20 s × 14 cycles; 72 °C for 1 min; held at 4 °C). Indexed To assess cytokine production by T lymphocytes, cells were (re-)stimulated for libraries were subjected a second double-sided size selection, and libraries were 5 h in serum-free CTL-Test™ PLUS Medium (#CTLT-010, ImmunoSpot) fi − − then quanti ed using on a Qubit 4 Fluorometer (Thermo Fisher). Quality was containing 20 ng mL 1 PMA and 1 µg mL 1 ionomycin together with BD assessed on a Bioanalyzer 2100, obtaining an average library size of 455 bp. GolgiPlug™ (1:100, #555029 from BD Biosciences). Afterwards, T cells were stained Libraries were finally diluted to 2 nM and clustered on a HiSeq2500 System with LIVE/DEAD™ Fixable Yellow dye (1:500, #L34959 from Thermo Fisher). Fc (Illumina) on high output mode at 10 pM on a pair-end read flow cell and receptors were blocked with the anti-mouse CD16/CD32 reagent Mouse BD Fc sequenced for 26 cycles on R1 (10X barcode and the UMIs), followed by 8 cycles of Block™ (1:200, clone 2.4G2, #553141 from BD Biosciences). Staining of surface I7 Index (sample Index), and 98 bases on R2 (transcript), with a coverage around markers was performed with the following fluorochrome-conjugated antibodies: 100 M reads per sample. Primary processing of sequencing images was done using anti-CD3-BV421 (1:100, clone 145-2C11, #562600 from BD Biosciences), anti- Illumina’s Real Time Analysis (RTA) software v. 3.4.4. CD8-FITC (1:400, clone 53-6.7, #553030 from BD Biosciences), and anti-CD4- PerCP-Cy5.5 (1;400, clone RM4-5, #45-0042-82 from Thermo Fisher). Cells were then fixed and permeabilized in Cytofix/Cytoperm™ buffer (#554714, BD Reproducibility and data management. Unless otherwise stated, all experiments Biosciences), and intracellular cytokine staining was performed with anti-IFN-γ- have been conducted in two independent instances with similar results. In each APC (1:100, clone XMG1.2, #554413 from BD Biosciences), anti-TNFα-APC-Cy7 individual experiment, samples were measured once. Transcriptomic and (1:200, clone MP6-XT22, #506307 from BioLegend), and anti-IL-2-PE (1:300, clone scRNAseq findings are from a single experiment involving the indicated number of JES6-%H4, #554428 from BD Biosciences). Gating procedure is exemplified in mice. Unless otherwise specified, Excel 2013 (Microsoft) and Prism v. 8.4 Supplementary Fig. 8. (GraphPad) were used for data management, graphing and statistical analyses.

NATURE COMMUNICATIONS | (2020) 11:3819 | https://doi.org/10.1038/s41467-020-17644-0 | www.nature.com/naturecommunications 15 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-17644-0

Illustrator 2020 (Adobe) and Photoshop 2020 (Adobe) were used for figure generated using ggpubr v. 0.2 and ggplot2 v. 3.1.1. Statistical assessments on dif- preparation. ferential gene expression and enrichment analysis were based on the MAST model70 and fast gene set enrichment analysis (FGSEA)71, respectively. Statistical significance on individual genes was performed by two-sided Wilcoxon test. Statistical analysis—tumor incidence and growth. Tumor surface was routinely evaluated by the ellipse area formula: (π × A × B)/4, where A and B are the largest and smallest lesion diameter, respectively. Statistical significance on growth curves Reporting summary. Further information on research design is available in the Nature was assessed by the TumGrowth software v. 158 or two-way ANOVA corrected for Research Reporting Summary linked to this article. row (time) and column (treatment) factors. Statistical significance on Kaplan Meyer curves (TFS, OS and TTD) was assessed by two-sided log-rank (Cox pro- Data availability portional hazards model) or one way-ANOVA plus Fisher LSD test. Transcriptomic data generated in the context of this study have deposited at Gene Expression Omnibus (GEO) and are publicly available under accession numbers Statistical analysis—in vitro, imaging and flow cytometry assays. Statistical GSE150921, GSE150966, GSE150967 and GSE151197. The Cancer Genome Atlas significance on RT-PCR, immunofluorescence microscopy, and flow cytometry (TCGA) breast cancer (BRCA) dataset is publicly available at https://xenabrowser.net/ (including cell death) data were assessed by one way-ANOVA plus Fisher LSD test datapages/?cohort=GDC%20TCGA%20Breast%20Cancer%20(BRCA) for comparisons involving more than two groups of samples or unpaired, two- &removeHub=https%3 A%2 F%2Fxena.treehouse.gi.ucsc.edu%3A443. The mm10 sided Student’s t test for comparisons involving only two groups of samples. mouse genome assembly and ENSEMBL Genes 100 can be accessed at https://useast. ensembl.org/Mus_musculus/Info/Index and https://www.ensembl.org/biomart/ Statistical analysis—bulk RNAseq. For data presented in Fig. 1, we used the martview/eb26fab860902330900c3b118e3cd906, respectively. GO terms, KEGG Affymetrix© Mouse Gene v. 1.0 ST Array. Data were normalized using Affymetrix© pathways, REACTOME pathways and Hallmarks terms are freely available to academic TAC v. 4.0.1. Analyses was performed within the R computational environment v. 3.4. users at http://geneontology.org/docs/downloads/, https://www.kegg.jp/kegg/download/, Specifically, genes were selected according to their differential expression by using https://reactome.org/download-data, and https://www.gsea-msigdb.org/gsea/msigdb/ limma v. 3.32.10 (ref. 59), using as thresholds |logFC| > 2 and p value < 0.05. GO collections.jsp#H, respectively. Additional data supporting the findings of this study are enrichment was analyzed using the DAVID Functional Annotation Tool v. 6.8 (ref. 60) available from the corresponding authors upon reasonable request. A reporting summary on differentially expressed genes. For comparing mouse RNAseq data with gene for this article is available as a Supplementary Information File. Unique materials used expression in human BC (data presented in Supplementary Fig. 1), we harnessed for this research are available from the corresponding authors upon reasonable microarray data from the TCGA database. In the R computational environment v. request. Source data are provided with this paper. 3.6.0, we renormalized mouse data (centered and reduced) in order to obtain similar distribution as human data. We then applied the function subtype.cluster of genefu v. Code availability 2.16.0 (ref. 61) based on previously published BC classification algorithms62,63.Other classification methods of the package genefu produced similar results. Heatmap with Custom code used for this research has been deposited at GitHub and is available at hierarchical clustering analysis was done using ComplexHeatmap v. 2.0.0 based on the https://github.com/nyuhuyang/scRNAseq-LorenzoBlood, and https://github.com/ Euclidean distance and ward.D2 clustering method. ENSEMBL Genes 100 was used to tonedivad/TumGrowth, or from the corresponding authors upon reasonable request. identified orthologous genes between human and mouse. For data presented in Fig. 6 and Supplementary Fig. 3, analysis was performed Received: 13 July 2019; Accepted: 10 July 2020; by GenoSplice technology (www.genosplice.com). Reads were mapped using STAR v. 2.4.0 (ref. 64) on the mm10 mouse genome assembly, followed by gene expression analysis65. In brief, for each gene present in Mouse FAST DB v. 2016_1 annotations, reads aligning on constitutive regions (that are not prone to alternative splicing) were counted. Based on these counts, normalization and differential gene expression were performed using DESeq2 v. 1.28.1 (ref. 66)onthe R computational environment (v.3.2.5). Only genes expressed in at least one of References experimental conditions were analyzed further. To this aim, genes were considered 1. Munoz, D. et al. Effects of screening and systemic adjuvant therapy as expressed if their rpkm value was greater than 93% of the background rpkm on ER-specific US breast cancer mortality. J. Natl Cancer Inst. 106, dju289 value based on intergenic regions. Results were considered statistically significant (2014). for p values ≤ 0.05 and fold changes ≥ 1.5. Gene modules were defined based on co- 2. Henry, N. L. et al. Role of Patient and Disease Factors in Adjuvant Systemic expression of genes within a given module. All genes within a given module had to Therapy Decision Making for Early-Stage, Operable Breast Cancer: American fi be correlated to each other. A minimum Pearson correlation coef cient of 0.7 (i.e., Society of Clinical Oncology Endorsement of Cancer Care Ontario Guideline only positive correlation) was required and a minimum of 50 genes was used to Recommendations. J. Clin. Oncol. 34, 2303–2311 (2016). fi de ne a module. Clustering and heatmaps were obtained using the dist and hclust 3. Shapiro, C. L. & Recht, A. Side effects of adjuvant treatment of breast cancer. functions of R, based on the Euclidean distance and Ward agglomeration method. N. Engl. J. Med. 344, 1997–2008 (2001). Analysis for enriched GO terms, KEGG pathways and REACTOME pathways were 4. Sharma, P. & Allison, J. P. The future of immune checkpoint therapy. Science performed using the DAVID Functional Annotation Tool v. 6.8. GO terms and 348,56–61 (2015). pathways were considered enriched if fold enrichment ≥ 2.0, uncorrected p value ≤ ≥ 5. Galluzzi, L., Chan, T. A., Kroemer, G., Wolchok, J. D. & Lopez-Soto, A. The 0.05, and minimum number of regulated genes in pathway/term 2.0. Analysis was hallmarks of successful anticancer immunotherapy. Sci. Transl. Med. 10, performed on all regulated genes, upregulated genes only, and downregulated genes eaat7807 (2018). only. A synthesis of these analyses was made to provide a single list of results. 6. Kroemer, G., Senovilla, L., Galluzzi, L., Andre, F. & Zitvogel, L. Natural and Analysis of the enriched GO pathways for commonly upregulated genes on both therapy-induced immunosurveillance in breast cancer. Nat. Med. 21, AT3 and M/D-driven tumors was performed using Ingenuity® Pathway Analysis 1128–1138 (2015). (IPA®) v. 52912811. 7. Schmid, P. et al. Atezolizumab plus nab-paclitaxel as first-line treatment for unresectable, locally advanced or metastatic triple-negative breast cancer Statistical analysis—single-cell RNASeq. The 10X Genomics Cell Ranger Single (IMpassion130): updated efficacy results from a randomised, double-blind, Cell Software Suite v2.1.0 was used to perform sample de-multiplexing, alignment placebo-controlled, phase 3 trial. Lancet Oncol. 21,44–59 (2020). (based on the mm10 mouse genome assembly), filtering, UMI counting, single-cell 8. Rugo, H. S. et al. Safety and antitumor activity of pembrolizumab in 3′end gene counting and performing quality control, using the manufacturer patients with estrogen receptor-positive/human epidermal growth factor parameters. Unsupervised cell clustering was carried out using Seurat v. 2.3.4 receptor 2-negative advanced breast cancer. Clin. Cancer Res. 24, 2804–2811 (ref. 67) in the R computational environment v. 3.5.3. Cells with <1000 genes or (2018). <3000 UMIs or >10% mitochondrial genes and genes detected in <3 cells were 9. Buque, A. & Galluzzi, L. Modeling tumor immunology and immunotherapy in excluded from the analysis. Gene expression raw counts were normalized following mice. Trends Cancer 4, 599–601 (2018). a global-scaling normalization method with a scale factor of 10,000 and a log 10. Aldaz, C. M., Liao, Q. Y., LaBate, M. & Johnston, D. A. Medroxyprogesterone transformation, using the Seurat NormalizeData function. The top 2775 highly acetate accelerates the development and increases the incidence of mouse variable genes were selected, followed by principal component analysis (PCA). Top mammary tumors induced by dimethylbenzanthracene. Carcinogenesis 17, 35 PCs were further used on harmony v. 0.1.0 (ref. 68) and remove the effect of – fi 2069 2072 (1996). confounding factors between two samples. The rst 35 harmony results were used 11. Chlebowski, R. T. et al. Influence of estrogen plus progestin on breast cancer to generate two-dimensional t-Distributed Stochastic Neighbor Embedding and mammography in healthy postmenopausal women: the Women’s Health (RunTSNE in Seurat, perplexity = 30) and unsupervised cell clustering by a shared Initiative Randomized Trial. JAMA 289, 3243–3253 (2003). nearest neighbor (FindClusters in Seurat, k.param = 30 and resolution = 0.6). A 12. Shen, J. et al. Dependence of cancer risk from environmental exposures on list of conserved cell-type specific genes was generated by FindAllMarkers function underlying genetic susceptibility: an illustration with polycyclic aromatic in Seurat. User defined genes were employed to identify cell types and confirmed by hydrocarbons and breast cancer. Br. J. Cancer 116, 1229–1233 (2017). SingleR v. 0.2.2 (ref. 69). Gene set enrichment analysis and bar charts were

16 NATURE COMMUNICATIONS | (2020) 11:3819 | https://doi.org/10.1038/s41467-020-17644-0 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-17644-0 ARTICLE

13. Billon-Gales, A. et al. Activation function 2 (AF2) of estrogen receptor-alpha 43. Scragg, R. et al. Monthly high-dose vitamin D supplementation and cancer is required for the atheroprotective action of estradiol but not to accelerate risk: a post hoc analysis of the vitamin D assessment randomized clinical trial. endothelial healing. Proc. Natl Acad. Sci. USA 108, 13311–13316 (2011). JAMA Oncol. 4, e182178 (2018). 14. Adlanmerini, M. et al. Mutation of the palmitoylation site of estrogen receptor 44. Manson, J. E. et al. Vitamin D supplements and prevention of cancer and alpha in vivo reveals tissue-specific roles for membrane versus nuclear actions. cardiovascular disease. N. Engl. J. Med. 380,33–44 (2019). Proc. Natl Acad. Sci. USA 111, E283–E290 (2014). 45. Elangovan, S. et al. The niacin/butyrate receptor GPR109A suppresses 15. Galluzzi, L., Buque, A., Kepp, O., Zitvogel, L. & Kroemer, G. Immunogenic mammary tumorigenesis by inhibiting cell survival. Cancer Res. 74, cell death in cancer and infectious disease. Nat. Rev. Immunol. 17,97–111 1166–1178 (2014). (2017). 46. Wise, A. et al. Molecular identification of high and low affinity receptors for 16. Galluzzi, L. et al. Consensus guidelines for the definition, detection and nicotinic acid. J. Biol. Chem. 278, 9869–9874 (2003). interpretation of immunogenic cell death. J. Immunother. Cancer 8, e000337 47. Shats, I. et al. Bacteria boost mammalian host NAD metabolism by (2020). engaging the deamidated biosynthesis pathway. Cell Metab. 31, 564–579 e567 17. Chlebowski, R. T. et al. Association of low-fat dietary pattern with breast (2020). cancer overall survival: a secondary analysis of the women’s health initiative 48. Wang, T., Cui, H., Ma, N. & Jiang, Y. Nicotinamide-mediated inhibition of randomized clinical trial. JAMA Oncol. 4, e181212 (2018). SIRT1 deacetylase is associated with the viability of cancer cells exposed to 18. Buck, M. D., Sowell, R. T., Kaech, S. M. & Pearce, E. L. Metabolic Instruction antitumor agents and apoptosis. Oncol. Lett. 6, 600–604 (2013). of Immunity. Cell 169, 570–586 (2017). 49. Hwang, E. S. & Song, S. B. Nicotinamide is an inhibitor of SIRT1 in vitro, but 19. Eisenberg, T. et al. Cardioprotection and lifespan extension by the natural can be a stimulator in cells. Cell Mol. Life Sci. 74, 3347–3362 (2017). polyamine spermidine. Nat. Med. 22, 1428–1438 (2016). 50. Limagne, E. et al. Sirtuin-1 activation controls tumor growth by 20. Menon, R. M. et al. Effect of the rate of niacin administration on the plasma impeding Th17 differentiation via STAT3 deacetylation. Cell Rep. 19, 746–759 and urine pharmacokinetics of niacin and its metabolites. J. Clin. Pharm. 47, (2017). 681–688 (2007). 51. Daenthanasanmak, A. et al. Targeting Sirt-1 controls GVHD by inhibiting T- 21. Engelhardt, J. J. et al. Marginating dendritic cells of the tumor cell allo-response and promoting Treg stability in mice. Blood 133, 266–279 microenvironment cross-present tumor antigens and stably engage tumor- (2019). specific T cells. Cancer Cell 21, 402–417 (2012). 52. Hogan, K. A., Chini, C. C. S. & Chini, E. N. The multi-faceted ecto-enzyme 22. Guy, C. T., Cardiff, R. D. & Muller, W. J. Induction of mammary tumors by CD38: roles in immunomodulation, cancer, aging, and metabolic diseases. expression of polyomavirus middle T oncogene: a transgenic mouse model for Front Immunol. 10, 1187 (2019). metastatic disease. Mol. Cell Biol. 12, 954–961 (1992). 53. Chabanon, R. M. et al. PARP inhibition enhances tumor cell-intrinsic 23. Kang, H. T. & Hwang, E. S. Nicotinamide enhances mitochondria quality immunity in ERCC1-deficient non-small cell lung cancer. J. Clin. Investig. 129, through autophagy activation in human cells. Aging Cell 8, 426–438 (2009). 1211–1228 (2019). 24. Qu, X. et al. Promotion of tumorigenesis by heterozygous disruption of the 54. Wang, Z. et al. Niraparib activates interferon signaling and potentiates anti- beclin 1 autophagy gene. J. Clin. Investig. 112, 1809–1820 (2003). PD-1 antibody efficacy in tumor models. Sci. Rep. 9, 1853 (2019). 25. Cicchini, M. et al. Autophagy regulator BECN1 suppresses mammary 55. Wickenhauser, C. et al. Multispectral fluorescence imaging allows for tumorigenesis driven by WNT1 activation and following parity. Autophagy distinctive topographic assessment and subclassification of tumor-infiltrating 10, 2036–2052 (2014). and surrounding immune cells. Methods Mol. Biol. 1913,13–31 (2019). 26. Galluzzi, L., Bravo-San Pedro, J. M., Levine, B., Green, D. R. & Kroemer, G. 56. Feng, Z. et al. Multispectral imaging of T and B cells in murine spleen and Pharmacological modulation of autophagy: therapeutic potential and tumor. J. Immunol. 196, 3943–3950 (2016). persisting obstacles. Nat. Rev. Drug Discov. 16, 487–511 (2017). 57. Galluzzi, L. et al. Guidelines for the use and interpretation of assays for 27. Berton, T. R. et al. Tumor formation in mice with conditional inactivation of monitoring cell death in higher eukaryotes. Cell Death Differ. 16, 1093–1107 Brca1 in epithelial tissues. Oncogene 22, 5415–5426 (2003). (2009). 28. De Giovanni, C. et al. Bioprofiling TS/A murine mammary cancer for a 58. Enot, D. P., Vacchelli, E., Jacquelot, N., Zitvogel, L. & Kroemer, G. functional precision experimental model. Cancers 11, 1889 (2019). TumGrowth: An open-access web tool for the statistical analysis of tumor 29. Stewart, T. J. & Abrams, S. I. Altered immune function during long-term host- growth curves. Oncoimmunology 7, e1462431 (2018). tumor interactions can be modulated to retard autochthonous neoplastic 59. Ritchie, M. E. et al. limma powers differential expression analyses for RNA- growth. J. Immunol. 179, 2851–2859 (2007). sequencing and microarray studies. Nucleic Acids Res. 43, e47 (2015). 30. Dikic, I. & Elazar, Z. Mechanism and medical implications of mammalian 60. Huang da, W., Sherman, B. T. & Lempicki, R. A. Systematic and integrative autophagy. Nat. Rev. Mol. Cell Biol. 19, 349–364 (2018). analysis of large gene lists using DAVID bioinformatics resources. Nat. Protoc. 31. Assarsson, E. et al. CD8+ T cells rapidly acquire NK1.1 and NK cell- 4,44–57 (2009). associated molecules upon stimulation in vitro and in vivo. J. Immunol. 165, 61. Gendoo, D. M. et al. Genefu: an R/Bioconductor package for computation of 3673–3679 (2000). gene expression-based signatures in breast cancer. Bioinformatics 32, 32. Abba, M. C. et al. DMBA induced mouse mammary tumors display high 1097–1099 (2016). incidence of activating Pik3caH1047 and loss of function Pten mutations. 62. Desmedt, C. et al. Biological processes associated with breast cancer clinical Oncotarget 7, 64289–64299 (2016). outcome depend on the molecular subtypes. Clin. Cancer Res. 14, 5158–5165 33. Shankaran, V. et al. IFNgamma and lymphocytes prevent primary tumour (2008). development and shape tumour immunogenicity. Nature 410, 1107–1111 63. Wirapati, P. et al. Meta-analysis of gene expression profiles in breast cancer: (2001). toward a unified understanding of breast cancer subtyping and prognosis 34. Swann, J. B. & Smyth, M. J. Immune surveillance of tumors. J. Clin. Investig. signatures. Breast Cancer Res. 10, R65 (2008). 117, 1137–1146 (2007). 64. Dobin, A. et al. STAR: ultrafast universal RNA-seq aligner. Bioinformatics 29, 35. Lopez-Soto, A., Gonzalez, S., Smyth, M. J. & Galluzzi, L. Control of Metastasis 15–21 (2013). by NK Cells. Cancer Cell 32, 135–154 (2017). 65. Noli, L., Capalbo, A., Ogilvie, C., Khalaf, Y. & Ilic, D. Discordant growth of 36. Andre, P. et al. Anti-NKG2A mAb Is a checkpoint inhibitor that promotes monozygotic twins starts at the blastocyst stage: a case study. Stem Cell Rep. 5, anti-tumor immunity by unleashing both T and NK cells. Cell 175, 1731–1743 946–953 (2015). e1713 (2018). 66. Love, M. I., Huber, W. & Anders, S. Moderated estimation of fold change and 37. Janssens, G. O. et al. Accelerated radiotherapy with carbogen and dispersion for RNA-seq data with DESeq2. Genome Biol. 15, 550 (2014). nicotinamide for laryngeal cancer: results of a phase III randomized trial. J. 67. Butler, A., Hoffman, P., Smibert, P., Papalexi, E. & Satija, R. Integrating single- Clin. Oncol. 30, 1777–1783 (2012). cell transcriptomic data across different conditions, technologies, and species. 38. Hoskin, P. J., Rojas, A. M., Bentzen, S. M. & Saunders, M. I. Radiotherapy with Nat. Biotechnol. 36, 411–420 (2018). concurrent carbogen and nicotinamide in bladder carcinoma. J. Clin. Oncol. 68. Korsunsky, I. et al. Fast, sensitive, and flexible integration of single cell data 28, 4912–4918 (2010). with Harmony. Preprint at https://www.biorxiv.org/content/10.1101/ 39. Chen, A. C. et al. A phase 3 randomized trial of nicotinamide for skin-cancer 461954v2 (2018). chemoprevention. N. Engl. J. Med. 373, 1618–1626 (2015). 69. Aran, D. et al. Reference-based analysis of lung single-cell sequencing reveals a 40. Minocha, R. et al. A Reduction in inflammatory macrophages may contribute transitional profibrotic macrophage. Nat. Immunol. 20, 163–172 (2019). to skin cancer chemoprevention by nicotinamide. J. Investig. Dermatol. 139, 70. Finak, G. et al. MAST: a flexible statistical framework for assessing 467–469 (2019). transcriptional changes and characterizing heterogeneity in single-cell RNA 41. Galbraith, A. R. et al. Chemoprevention of Lung Carcinogenesis by Dietary sequencing data. Genome Biol. 16, 278 (2015). Nicotinamide and Inhaled Budesonide. Cancer Prev. Res. 12,69–78 (2019). 71. Korotkevich, G., Sukhov, V. & Sergushichev, A. Fast gene set enrichment 42. Zhang, S. M. et al. Effect of combined folic acid, vitamin B6, and vitamin B12 analysis. Preprint at https://www.biorxiv.org/content/10.1101/060012v2 on cancer risk in women: a randomized trial. JAMA 300, 2012–2021 (2008). (2019).

NATURE COMMUNICATIONS | (2020) 11:3819 | https://doi.org/10.1038/s41467-020-17644-0 | www.nature.com/naturecommunications 17 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-17644-0

Acknowledgements project and analyses, designed data collection protocols, designed and conducted We are grateful to F. Lenfant (INSERM U1048, Toulouse, France) and J. Penninger the analyses, interpreted the results, wrote the manuscript and supervised figure (Institute of Molecular Biotechnology, Vienna, Austria) for kindly providing us with preparation. mice bearing genetic alterations in Esr1 and Atg5, respectively. We are indebted to B. Levine (UT Southwestern, Dallas, TX, USA) and Dr. M.F. Krummel (UCSF, San + − Competing interests Francisco, CA, USA) for kindly providing us with Becn1 / and MMTV-PyMT mice, S.C.F. reports funding for clinical trials from Bristol Myers Squibb, Merck and Varian, respectively. We thank Genom’IC (Cochin Institute, Paris, France) and GenoSplice and speaker and/or advisory honoraria from Astra Zeneca, Bayer, Bristol Myers Squibb, (Paris, France) for help with RNA-Seq analysis. We thank S. Hoda (Weill Cornell Eisai, Elekta, EMD Serono/Merck, GlaxoSmithKline, Janssen, MedImmune, Merck US, Medicine, New York, NY, US) for providing human breast cancer microphotographs, Regeneron, Varian, and ViewRay. O.E. reports research funding from Eli Lilly, Janssen, G. Inghirami (Weill Cornell Medicine, New York, NY, US) for pathology assessments, Sanofi, equities/co-founder role in Volastra Therapeutics and OneThree Biotech, and G. Petroni (Weill Cornell Medicine, New York, NY, US) for critical reading and help equities/advisory role in Owkin, Freenome, Genetic Intelligence and Acuamark DX. with the finalization of the paper, and the Proteomics and Metabolomics Core Facility of F. André reports research funding and speaker/advisory honoraria from Roche, Astra- Weill Cornell Medicine (New York, NY, US) for NAM quantification. M.P.L. is sup- Zeneca, Daiichi Sankyo, Pfizer, Novartis, and Eli Lilly. L.Z. reports research funding from ported by Fondation pour la Recherche Médicale (FRM FDT201904008383). J.H. is Bristol Myers Squibb, Roche, Glaxo Smyth Kline, Lytix Pharma, Incyte, Merus, Tusk and supported by the Foundation Philanthropia. B.S. is supported from grants by the German Pileje (completed), and from Innovate Pharma, Kaleido, Transgene, Elior, Carrefour Cancer Aid (#70113311 and #34102524). G.K. is supported by the Ligue contre le Cancer (ongoing), consulting/advisory honoraria from Transgene, EpiVax, Lytix Biopharma, and (équipe labellisée); Agence National de la Recherche (ANR) – Projets blancs; ANR under co-founder role in everImmune. S.D. reports funding for clinical trials from AstraZeneca, the frame of E-Rare-2, the ERA-Net for Research on Rare Diseases; Association pour la Roche, Bristol Myers Squibb, Pfizer, Novartis, Amgen, Sanofi, GE Healthcare, Merck recherche sur le cancer (ARC); Cancéropôle Ile-de-France; Chancelerie des universités Sharp & Dohme, Eli Lilly, Puma, Myriad, Orion, Genomic Health, Pierre Fabre and de Paris (Legs Poix), Fondation pour la Recherche Médicale (FRM); a donation by Elior; Servier, non-financial support from AstraZeneca, Roche and Pfizer. G.K. reports research European Research Area Network on Cardiovascular Diseases (ERA-CVD, MINO- funding from Bayer Healthcare, Genentech, Glaxo Smyth Kline, Institut Mérieux, Lytix TAUR); Gustave Roussy Odyssea, the European Union Horizon 2020 Project Onco- Pharma, PharmaMar, Sotio and Vasculox, consulting/advisory honoraria from Bristol biome; Fondation Carrefour; High-end Foreign Expert Program in China Myers Squibb Foundation France, co-founder role of everImmune, Samsara therapeutics (GDW20171100085 and GDW20181100051), Institut National du Cancer (INCa); and Therafast Bio. L.G. reports research funding from Lytix, and Phosplatin (completed), Inserm (HTE); Institut Universitaire de France; LeDucq Foundation; the LabEx consulting/advisory honoraria from Boehringer Ingelheim, AstraZeneca, OmniSEQ, The Immuno-Oncology; the RHU Torino Lumière; the Seerave Foundation; the SIRIC Longevity Labs, Inzen, and the Luke Heller TECPR2 Foundation. All other authors have Stratified Oncology Cell DNA Repair and Tumor Immune Elimination (SOCRATE); and no conflicts of interest to declare. the SIRIC Cancer Research and Personalized Medicine (CARPEM). The LG lab is supported by a Breakthrough Level 2 grant from the US Department of Defense (DoD), Breast Cancer Research Program (BRCP) (#BC180476P1), by the 2019 Laura Ziskin Additional information Prize in Translational Research (#ZP-6177, PI: Formenti) from the Stand Up to Cancer Supplementary information is available for this paper at https://doi.org/10.1038/s41467- (SU2C), by a Mantle Cell Lymphoma Research Initiative (MCL-RI, PI: Chen-Kiang) 020-17644-0. grant from the Leukemia and Lymphoma Society (LLS), by a startup grant from the Dept. of Radiation Oncology at Weill Cornell Medicine (New York, US), by a Rapid Correspondence and requests for materials should be addressed to G.K. or L.G. Response Grant from the Functional Genomics Initiative (New York, US), by industrial collaborations with Lytix (Oslo, Norway) and Phosplatin (New York, US), and by Peer review information Nature Communications thanks the anonymous reviewer(s) for donations from Phosplatin (New York, US), the Luke Heller TECPR2 Foundation their contribution to the peer review of this work. Peer reviewer reports are available. (Boston, US) and Sotio a.s. (Prague, Czech Republic). Reprints and permission information is available at http://www.nature.com/reprints

Author contributions Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in fi A. Buqué and N.B. performed the majority of experimental procedures, collected results, published maps and institutional af liations. analyzed data, wrote technical sections of the manuscript, and prepared figures for publication. M.P.-L., K.I., J.H., and F. Aranda provided major support to tumor estab- lishment, survival studies and sample collection. M.P.-L. helped with preparation of Open Access This article is licensed under a Creative Commons samples for RNAseq and metabolomics. M.P.-L., J.G.P., S.L., L.M., and J.F. performed Attribution 4.0 International License, which permits use, sharing, flow cytometry studies and analyzed data. T.Y. performed flow cytometry experiments adaptation, distribution and reproduction in any medium or format, as long as you give − − and generated Atg7 / TSA cells. A.S. performed biochemical and cellular studies on appropriate credit to the original author(s) and the source, provide a link to the Creative autophagy-competent and—deficient cells, as well as immunofluorescence microscopy Commons license, and indicate if changes were made. The images or other third party assays for ER and VIM expression. S.D. performed metabolic studies. A. Boissonnas material in this article are included in the article’s Creative Commons license, unless tested the effect of NAM on MMTV-PyMT mice. L.S. provided technical support to indicated otherwise in a credit line to the material. If material is not included in the model establishment. D.E., M.H., M.K., and G.S. performed statistical and bioinformatic article’s Creative Commons license and your intended use is not permitted by statutory analyses on RNAseq data. Y.H. analyzed single-cell RNAseq data. C.M. performed regulation or exceeds the permitted use, you will need to obtain permission directly from fl fi immuno uorescence microscopy assays for immune in ltration under supervision the copyright holder. To view a copy of this license, visit http://creativecommons.org/ from B.S. S.C.F., R.S., F. André, and S.D. provided infrastructure and guided transla- licenses/by/4.0/. tional/clinical relevance. O.E. supervised the analysis of single-cell RNAseq and provided infrastructure. L.Z. conceived the project and provided infrastructure and guided translational/clinical relevance. L.G. and G.K. conceived and directed the © The Author(s) 2020

18 NATURE COMMUNICATIONS | (2020) 11:3819 | https://doi.org/10.1038/s41467-020-17644-0 | www.nature.com/naturecommunications