ARTICLES https://doi.org/10.1038/s41591-019-0566-4
Stress–glucocorticoid–TSC22D3 axis compromises therapy-induced antitumor immunity
Heng Yang1,2,22, Lin Xia1,2,22, Jian Chen3,22, Shuqing Zhang1,2, Vincent Martin4, Qingqing Li1,2, Shangqing Lin1,2, Jinfeng Chen1,2, Joseph Calmette5, Min Lu! !6, Lingyi Fu7, Jie Yang7, Zhizhong Pan7, Kuai Yu7, Jingjing He7, Eric Morand! !8, Géraldine Schlecht-Louf! !5, Roman Krzysiek5,9, Laurence Zitvogel1,2,10,11,12, Boxi Kang! !13, Zeming Zhang! !13, Andrew Leader14, Penghui Zhou7, Laurence Lanfumey! !4, Minxin Shi3, Guido Kroemer! !1,2,15,16,17,18,19,20,21,23* and Yuting Ma! !1,2,23*
Psychological distress has long been suspected to influence cancer incidence and mortality. It remains largely unknown whether and how stress affects the efficacy of anticancer therapies. We observed that social defeat caused anxiety-like behaviors in mice and dampened therapeutic responses against carcinogen-induced neoplasias and transplantable tumors. Stress elevated plasma corticosterone and upregulated the expression of glucocorticoid-inducible factor Tsc22d3, which blocked type I inter- feron (IFN) responses in dendritic cell (DC) and IFN-γ+ T cell activation. Similarly, close correlations were discovered among plasma cortisol levels, TSC22D3 expression in circulating leukocytes and negative mood in patients with cancer. In murine mod- els, exogenous glucocorticoid injection, or enforced expression of Tsc22d3 in DC was sufficient to abolish therapeutic control of tumors. Administration of a glucocorticoid receptor antagonist or DC-specific Tsc22d3 deletion reversed the negative impact of stress or glucocorticoid supplementation on therapeutic outcomes. Altogether, these results indicate that stress-induced glucocorticoid surge and Tsc22d3 upregulation can subvert therapy-induced anticancer immunosurveillance.
atients with cancer often rationalize their diseases as a conse- psychosocial stress on the outcome of cancer treatment has not been quence of adverse life experiences, which contrasts with large investigated in detail. prospective studies that have failed to establish such a correla- Before the advent of immunotherapy, cancer was usually treated P1 tion . In sharp opposition, there is overwhelming epidemiological by surgery, chemotherapy, radiotherapy and targeted therapies. The evidence that dietary factors, in particular overweight and obesity, goal of this treatment was reduction of the proliferation of malig- can influence the morbidity and mortality of a number of distinct nant cells (cytostasis) or elimination of them by activating cell-death cancers2, a fact that is often denied at the individual, subjective level. programs (cytotoxicity). The overarching logic of this approach was Beyond the non-objectivity in the patients’ appreciation of their consideration of neoplasm as a disease caused solely by malignant mental and bodily selves, cancer is a highly heterogeneous disease cells, meaning that the therapeutic agents needed to target cancer in which the (epi)genetic makeup of malignant cells as well as the cells themselves5. In this perspective, which simplifies cancer to a particularities of the tumor microenvironment make each individ- cell-biology problem, there is no space to explain how psychosocial ual case unique3. For these reasons, psychological factors, including and neuroendocrine factors might affect the efficacy of chemother- mental stress, may be difficult to relate to the incidence, progression apy or radiotherapy. and therapeutic response of malignant diseases. Although one large Recent work, however, has led a radical change in the view on meta-analysis (on 163,363 individuals) revealed that self-reported how therapeutic success (that is, long-term control, or even a cure, psychological-distress scores are epidemiologically linked to can- of the disease) may be achieved by cancer therapies. Indeed, when cer-related mortality (in particular to colorectal, prostate, pancre- the effects of treatment last beyond treatment discontinuation, atic and esophageal cancer, as well as to leukemia)4, the impact of these might be attributed to the induction of an anticancer immune
1Center for Systems Medicine, Institute of Basic Medical Sciences, Chinese Academy of Medical Sciences & Peking Union Medical College, Beijing, China. 2Suzhou Institute of Systems Medicine, Suzhou, Jiangsu, China. 3The Affiliated Tumor Hospital of Nantong University, Nantong Tumor Hospital, Nantong, Jiangsu, China. 4INSERM UMR S894, Centre de Psychiatrie et Neuroscience, Université Paris Descartes, Paris, France. 5UMR996, Inflammation, Chemokines and Immunopathology, INSERM, Université Paris-Sud, Université Paris-Saclay, Clamart, France. 6State Key Laboratory of Medical Genomics, Shanghai Institute of Hematology, Rui Jin Hospital affiliated to Shanghai Jiaotong University School of Medicine, Shanghai, China. 7State Key Laboratory of Oncology in Southern China, Collaborative Innovation Center for Cancer Medicine, Sun Yat-Sen University Cancer Center, Guangzhou, China. 8Clinical Sciences, Monash University, Monash Medical Centre, Clayton, Victoria, Australia. 9Laboratoire d’Immunologie Biologique, CHU du Kremlin Bicêtre, AP-HP, Paris, France. 10INSERM U1015, Gustave Roussy Cancer Campus (GRCC), Villejuif, France. 11Université Paris-Saclay, Villejuif, France. 12Center of Clinical Investigations in Biotherapies of Cancer (CICBT) 1428, Villejuif, France. 13BIOPIC, Beijing Advanced Innovation Centre for Genomics, and School of Life Sciences, Peking University, Beijing, China. 14Department of Oncological Sciences, Icahn School of Medicine at Mount Sinai, New York, NY, USA. 15Université de Paris, Paris, France. 16Equipe 11 labellisée Ligue Nationale contre le Cancer, Centre de Recherche des Cordeliers, Paris, France. 17Institut National de la Santé et de la Recherche Médicale U1138, Paris, France. 18Sorbonne Université, Paris, France. 19Metabolomics and Cell Biology Platforms, Gustave Roussy Cancer Campus, Villejuif, France. 20Pôle de Biologie, Hôpital Européen Georges Pompidou, Assistance Publique – Hopitaux de Paris, Paris, France. 21Department of Women’s and Children’s Health, Karolinska University Hospital, Stockholm, Sweden. 22These authors contributed equally: Heng Yang, Lin Xia, Jian Chen. 23These authors jointly supervised this work: Guido Kroemer and Yuting Ma. *e-mail: [email protected]; [email protected]
1428 NATURE MEDICINE | VOL 25 | SEPTEMBER 2019 | 1428–1441 | www.nature.com/naturemedicine NATURE MEDICINE ARTICLES response, meaning that therapeutic intervention with drugs or activities in the sucrose splash test (Extended Data Fig. 1i) and irradiation has converted the tumor into its own vaccine6. This avoidance of social interactions (Extended Data Fig. 1j–l). We has been exemplified experimentally and clinically for several suc- determined the effect of SD on urethane-induced non-small-cell cessful chemotherapeutic agents, such as anthracyclines and oxali- lung cancer (NSCLC) in mice, in which we scored the spontaneous platin (OXA), that are widely used for the treatment of mammary development of tumors, as well as the response to chemotherapy and colorectal carcinomas, respectively. Anthracyclines and OXA with OXA, an ICD inducer (Fig. 1d), using micro-CT. SD in mice stimulate a type of cancer-cell death that is immunogenic owing to caused an increase in the number and tumor burden (tumor size in danger-associated molecular patterns (DAMPs) that are released mm2) of lung neoplasias relative to non-stressed controls. Although from malignant cells or are exposed on their surface to attract DC OXA-based chemotherapy reduced tumor mass and mean tumor precursors into the tumor bed, trigger the transfer of portions of size in unstressed mice, this therapeutic effect was largely compro- the malignant cell into an immature DC, stimulate DC maturation mised in SD-conditioned mice (Fig. 1e–g). Moreover, histopatho- and finally induce cross-presentation of tumor-associated antigens logical examination suggested that SD facilitated the formation to cytotoxic T lymphocytes (CTLs). Such tumor-antigen-specific of adenoma and adenocarcinoma and diminished the capacity of CTLs produce IFN-γ and then mediate the long-term effects of che- OXA to eradicate adenocarcinomas (Fig. 1h,i). We also explored the motherapy with respect to tumor-growth retardation7–9. impact of SD on colorectal cancers induced by a combination of Immunity is organized in a systemic rather than local fashion azoxymethane (AOM), a genotoxic carcinogen, and dextran sodium (the ‘immune system’) and is known to be profoundly influenced by sulfate (DSS), a pro-inflammatory and colitis-inducing agent neuroendocrine factors, including multiple neurotransmitters, neu- (Fig. 1j). SD consistently increased the number and size of colorectal ropeptides and hormones (such as glucocorticoids that increase in tumors. Moreover, OXA-based chemotherapy reduced the number the context of psychological distress)10–12, establishing functionally of detectable neoplasias in control mice, yet failed to do so in the SD important neuro–immune interactions13,14. Given the importance of group (Fig. 1k–m). It is noteworthy that SD procedures caused sur- immune effectors, such as DCs and CTLs, for the efficacy of immu- vival disadvantage in mice that received either PBS or OXA, indi- nogenic cell death (ICD)-inducing chemotherapies, this creates a cating that they had increased vulnerability to colon carcinogenesis possible mechanistic link between mental factors and therapeutic and/or ineffectiveness of chemotherapy (Extended Data Fig. 1m). outcomes. With this idea as a basis, we investigated a mouse model However, no baseline differences in the number and size of neopla- of stress, in which social defeat (SD) causes a long-term state of anx- sias were observed between the SD and Ctrl groups on day 55, when iety and/or depression, to explore the possible effect of psychologi- the OXA regimen started (Extended Data Fig. 1n,o). The 10-d SD cal distress on patient outcome following cancer treatment. procedure slightly increased the food and water consumption, but not body weight, of the mice (Extended Data Fig. 1p,r). Results Next, we recurred to transplantable MCA205 fibrosarcomas and Social defeat negatively affects cancer therapeutic outcomes and TC-1 NSCLC models to investigate the impact of SD on chemo- immunosurveillance. To investigate the potential impact of stress therapy-induced anticancer immunity (Extended Data Fig. 2a). In on cancer development and therapeutic efficacy, we took advantage the control group, the progression of MCA205 and TC-1 tumors of SD15 and acute restraint (AR)16 models (Fig. 1a and Extended was reduced in response to the ICD inducer mitoxantrone (MTX), Data Fig. 1a,b). Repeated exposure of male C57BL/6 mice to aggres- but such a therapeutic response was absent in SD mice (Fig. 2a,b sive CD-1 males causes anxiety-like behavior, which manifests as and Extended Data Fig. 2b,c). The subcutaneous injection of a pro- reduced exploration of the light compartment in the light–dark phylactic vaccine composed of dying MCA205 cells (which were box test (Fig. 1b and Extended Data Fig. 1c–e), diminished total pretreated with MTX ex vivo)17 inhibited the progression of live locomotion and exploration in the central area of the open-field MCA205 cells injected 10 d later into control mice. This tumor-pre- test (Fig. 1c and Extended Data Fig. 1f–h), reduced grooming ventive effect was significantly reduced by prior SD conditioning
Fig. 1 | SD accelerates tumor progression and hampers chemotherapeutic responsiveness. a, A schematic showing the repeated SD procedures. b, Typical locomotion tracks (green lines) of control (Ctrl) and mice subjected to SD in light–dark box tests (left). Locomotion (quantified as length of the track) in the light box was compared between Ctrl (n = 12 mice) and SD (n = 13 mice) groups (right). A two-tailed Mann–Whitney U test was used for statistical analysis (P < 0.0001). A representative result from five independent experiments is shown. c, Typical locomotion tracks of Ctrl and SD mice in open-field tests (left). Total locomotion (length of the track) was compared between Ctrl (n = 10 mice) and SD groups (n = 18 mice) (right). A two-tailed Mann–Whitney U test was used for statistical analysis (P < 0.0001). A representative result from five independent experiments is shown. d, A schematic of the timing of SD, urethane injections, behavioral tests, OXA-based chemotherapy, micro-CT and stereomicroscope examination in the primary lung cancer model. i.p., intraperitoneal injection. e, Typical images of neoplasias (marked by yellow arrows) on lung lobes from the indicated groups on day 120 after the first urethane injection. A representative result from two independent experiments is shown. f–g, The total size of neoplasias (f) and the numbers of large (>1 mm2), mid-size (0.5–1 mm2) and small (<0.5 mm2) neoplastic lesions (g) were compared between Ctrl and SD groups, with or without OXA treatment (n = 5 mice for SD group, n = 6 mice for all other groups). A two-tailed Mann–Whitney U test was used for statistical analyses (P values are shown between indicated groups). In g, the blue and black lines indicate significant changes in small and medium neoplasias, respectively. h, Typical images of lung adenoma and adenocarcinoma with H&E-staining-based histopathology analysis. A representative result from two independent experiments is shown. i, The number of adenomas and adenocarcinomas evaluated with paraformaldehyde-fixed, frozen lung sections embedded in optimum cutting temperature compound were compared between Ctrl and SD groups that received either solvent or OXA (n = 5 mice for SD group, n = 6 mice for all the other groups). Two coronal sections containing five pulmonary lobes from each mouse near the maximal diameters (i.e., transversal surface of lung lobes) were assessed. A two-tailed Mann–Whitney U test was used for statistical analysis (P values are shown between the indicated groups). j, A schematic of the timing of SD, administration of AOM and DSS, behavioral tests, OXA-based chemotherapy, micro-CT and stereomicroscope examination in the primary colorectal cancer model. p.o., per os. k, Typical images of colons from Ctrl and SD groups on day 90 following SD. A representative result from two independent experiments is shown. l,m, Total size of neoplasias (l), and the numbers of large (>7.5 mm2), mid-size (2.5–7.49 mm2) and small (<2.49 mm2) malignant lesions in the colon and rectum (m) were compared between Ctrl and SD groups on day 90, with or without OXA-based chemotherapy (n = 10 mice for Ctrl, n = 9 mice for SD, n = 12 mice for the other groups). A two-tailed Mann–Whitney U test was used for statistical analysis (P values are shown between the indicated groups). Red and black lines indicate statistics of large and middle neoplasias. All data are shown as mean ± s.e.m. *P < 0.05, **P < 0.01, ***P < 0.001; ns, not significant.
NATURE MEDICINE | VOL 25 | SEPTEMBER 2019 | 1428–1441 | www.nature.com/naturemedicine 1429 ARTICLES NATURE MEDICINE
(Fig. 2c and Extended Data Fig. 2d,e). Moreover, tumor-growth chemotherapy, as well as perturb responses to prophylactic tumor retardation by programmed death-1 (PD-1) blockade was less vaccination and PD-1-targeted immunotherapy. efficient in SD than in Ctrl mice (Fig. 2d and Extended Data Fig. 2f,g). Of note, similar to SD, three rounds of AR also compromised Social-defeat-induced alterations in the cancer immune contex- tumor-growth reduction by MTX in the MCA205 and TC-1 models ture. To investigate the potential immunosuppressive impact of social (Fig. 2e,f and Extended Data Fig. 2h–j). stress, we analyzed the immune infiltrates of MCA205 cancers that Altogether, these results indicate that SD- and AR-induced were developing in control and SD mice at 2 d after chemotherapy. stress negatively affect the therapeutic response to immunogenic Cytofluorometric analyses revealed that SD pretreatment reduced the
ab c Ctrl SD Light–dark box Open field P < 0.0001 Ctrl SD P < 0.0001 4,000 CD-1 3,000 *** ***
3,000 2,000
2,000 10 min C57BL/6 1,000
d 1,000 Total locomotion (cm) Light box locomotion (cm) SD 10 24 h 0 0 Light Dark 15 cm Ctrl SD Ctrl SD box box 20 cm d efCtrl SD ** P = 0.0022 Urethane-induced lung cancer 8 ** P = 0.0087 ns P = 0.1255 SD ) ** P = 0.0087 2 Urethane i.p. 6 Tumor PBS Behavioral test OXA i.p. 4 Micro-CT Stereomicroscope 2 Total tumor size (mm OXA
0 10 20 30 40 50 60 70 80 90 0 100 110 120 Ctrl SD Ctrl SD Days 10 mm OXA OXA g >1 mm2 hi j 0.5–1 mm2 Ctrl 2 12 SD 12 <0.5 mm *** P = 0.0009 AOM–DSS-induced colon cancer Ctrl OXA ** P = 0.0051 * P = 0.0325 SD OXA SD ** P = 0.0058 ** P = 0.0022 9 9 AOM i.p.
Adenoma ** * P = 0.0025 P = 0.042 Behavioral test ns P = 0.0686 6 6 DSS p.o. * * 200 m P = 0.048 µ P = 0.0169 OXA i.p. No. of neoplasias No. of neoplasia 3 3 Stereomicroscope
0 0 0102030405060708090 Ctrl SD Ctrl SD Adenocarcinoma Adenomas Adenocarcinomas Days OXA OXA klm >7.50 mm2 Ctrl SD ns P = 0.9256 2.50–7.49 mm2 300 75 * P = 0.0338 <2.49 mm2 ns P = 0.7775 ** P = 0.0056 * P = 0.0152
) * P = 0.0345
2 60 a
S ** P = 0.0024 200 PB 45
30 100 No. of neoplasi 15 Total tumor size (mm OXA 0 0 Ctrl SD Ctrl SD Ctrl SD Ctrl SD 20 mm OXA OXA OXA OXA
1430 NATURE MEDICINE | VOL 25 | SEPTEMBER 2019 | 1428–1441 | www.nature.com/naturemedicine NATURE MEDICINE ARTICLES frequency of T lymphocytes and neutrophils, whereas it increased immunosuppression, we investigated the impact of SD on neuroen- the density of macrophages without interfering with the abundance docrine stress mediators, such as corticosterone, serotonin and nor- of tumor-infiltrating dendritic cells (TIDCs) (Fig. 2g and Extended epinephrine. Plasma corticosterone was notably higher in SD mice Data Fig. 2k). Moreover, SD reorchestrated T cell differentiation in than in unstressed controls. Interestingly, elevated corticosterone the tumor microenvironment by reducing the expression of T-box 21 concentrations correlated with reduced locomotion (characteristic (Tbx21) and GATA binding protein 3 (Gata3), which encode type 1 T of SD-induced anxiety) at the individual (mouse per mouse) level helper cell (TH1)- and TH2-restricted transcription factors, respectively. (Fig. 3a,b). AR merely caused a transient increase in corticosterone The transcription of RAR-related orphan receptor gamma (Rorc) and levels, whereas SD induced a durable augmentation in the gluco- forkhead box P3 (Foxp3), which are master regulators of TH17 and corticoid tonus (Extended Data Fig. 3a,b). In contrast to corticos- regulatory T cell (Treg) differentiation, was unaltered. In line with the terone, plasma serotonin and norepinephrine concentrations were role that Tbx21 has in driving IFN-γ-dependent TH1 responses, SD decreased after SD procedures, and these changes were also long- inhibited interferon gamma (Ifng) expression in tumor-infiltrating lasting. The level of circulating serotonin, but not norepinephrine, in T cells (Fig. 2h). Accordingly, SD preconditioning significantly ham- SD and control mice positively correlated with their locomotion in pered IFN-γ secretion by T cells in transplanted MCA205 fibrosar- behavioral assessments (Fig. 3c–f and Extended Data Fig. 3c,d). We comas after MTX treatment (Fig. 2i), and it reduced the density of also noticed that the increment of plasma corticosterone was cou- IFN-γ-secreting CD3+ T cells in urethane-induced primary NSCLCs pled to an initial elevation of adenocorticotropic hormone (ACTH) after OXA therapy (Fig. 2j). In accordance, the plasma concentration levels during SD (Fig. 3g). SD-induced stress compromised the effi- of IFN-γ was elevated in control mice 7 d after chemotherapy, but this cacy of chemotherapy and immunotherapy, and this effect was abol- did not happen in the SD group (Fig. 2k). As a further indication of ished by administration of a glucocorticoid receptor antagonist18,19 systemic immune suppression, blood plasma collected from tumor- mifepristone (MIFE). Thus, repeated MIFE injections restored the bearing SD mice 48 h after MTX chemotherapy exhibited a relative capacity of MTX to inhibit tumor outgrowth, relieved anxiety-like deficiency in 46 circulating factors, including cytokines and chemo- behavior and the stress-triggered surge in corticosterone levels kines, such as interferon (IFN)-β, interleukin (IL)-15, IL-23A, C-X-C in SD mice (Fig. 3h and Extended Data Fig. 3e–g). Additionally, motif chemokine 10 (CXCL10), CXCL1 and CXCL9, as compared MIFE improved the outcome of PD-1 blockade and reestablished with plasma from their non-stressed counterparts (Fig. 2l). the efficacy of anticancer vaccination with dying tumor cells Altogether, these results demonstrate that stress can cause intra- (Fig. 3i,j and Extended Data Fig. 3h,i). This contrasts with the failure tumoral and systemic immunosuppression, thereby compromising of fluoxetine (FLU) and reboxetine (REB), which selectively inhibit therapy-induced anticancer immune responses. the reuptake of serotonin20,21 and norepinephrine22,23 respectively, to restore MTX-induced tumor-growth retardation (Extended Data Stress-induced glucocorticoid tonus subverts immunosur- Fig. 3j,k). Although FLU and REB did achieve an elevation of sero- veillance. To establish a mechanistic link between SD and tonin and norepinephrine levels, respectively, both treatments were
Fig. 2 | SD impairs the efficacy of anticancer therapies and induces immunosuppression. a,b, MCA205 fibrosarcoma cells (a) or TC-1 lung cancer cells (b) were inoculated in Ctrl or SD-preconditioned mice, and the tumor-retarding effect of MTX-based chemotherapy was compared between groups. For the MCA205 model, n = 5 mice for both Ctrl PBS and Ctrl MTX, n = 9 mice for SD PBS, n = 7 mice for SD MTX. A representative result from two independent experiments is shown in a. For the TC-1 model, n = 10 mice for Ctrl PBS, n = 11 mice for Ctrl MTX, n = 13 mice for SD PBS, n = 10 mice for SD MTX. Results from two independent experiments are combined and shown in b. Area under the curve was calculated for each mouse in all groups. A two-tailed Mann–Whitney U test was used for statistical analyses (P values are shown between the indicated groups). c, MCA205 tumor cells were pretreated with MTX and used as prophylactic vaccine (VAC) in Ctrl and SD mice (n = 6 for Ctrl PBS and Ctrl VAC, n = 5 for SD PBS, n = 7 for SD VAC). Tumor growth was monitored after re-challenging these mice with live MCA205 cells. The number of tumor-free (TF) mice is shown on each graph. A representative result from two independent experiments is shown. The area under the curve was calculated for each mouse in all groups. A two-tailed Mann–Whitney U test was used for statistical analyses (P values are shown between the indicated groups). d, The efficacy of immune checkpoint blockade with PD-1 mAb was compared in Ctrl and SD mice. Isotype control antibody (ISO) was used as control (n = 9 for Ctrl ISO, n = 8 for SD ISO and SD anti-PD-1, n = 10 for Ctrl anti-PD-1). The area under the curve was calculated for each mouse in all groups. A two-tailed Mann–Whitney U test was used for statistical analyses (P values are shown between the indicated groups). e,f, The tumor-growth-inhibition effect of MTX was compared between Ctrl and AR mice engrafted with MCA205 (n = 5 for AR MTX, n = 6 for all the other groups) (e) or TC-1 (n = 8 for Ctrl MTX, n = 7 for all the other groups) (f) tumors. A representative result from two independent experiments is shown for each cancer model. The area under the curve was calculated for each mouse in all groups. A two-tailed Mann–Whitney U test was used for statistical analyses (P values are shown between the indicated groups). g, FACS analyses of the proportions of major immune cell populations in MCA205 tumors, which were taken from Ctrl or SD mice 2 d after MTX (n = 21 for Ctrl MTX, n = 12 for SD MTX). Results from two independent experiments are combined and shown. A two-tailed Mann–Whitney U test was used for statistical analyses (P values are shown between the indicated groups). h, The relative expressions of genes encoding TH1-, TH2-, TH17-, and Treg-restricted transcriptional factors and the TH1 effector molecule IFN-γ in tumor-infiltrating T cells were quantified using the house-keeping gene Ppia (n = 6 mice for all groups). A representative result from two independent experiments is shown for each cancer model. A two-tailed Mann–Whitney U test was used for statistical analyses (P values are shown between the indicated groups). i, IFN-γ secretion by tumor-infiltrating T cells was quantified using ELISPOT and compared between Ctrl and SD mice 7 d after treatment with MTX or PBS (n = 9 for SD MTX, n = 11 for all the other groups). Results from two independent experiments are combined and shown. A two-tailed Mann–Whitney U test was used for statistical analyses (P values are shown between the indicated groups). j, Quantification of CD3+IFN-γ+ T cells in lung lobes taken from urethane-induced primary lung cancer models (n = 5 mice for SD group, n = 6 mice for all the other groups). Three coronal sections containing five pulmonary lobes from each mouse near the maximal diameters were quantified. Each dot represents one view field. A representative result from two independent experiments is shown. Typical images (left) and results from statistical analysis with the two-tailed Mann–Whitney U test (right) are shown. k, The plasma concentrations of IFN-γ were detected by ELISA and compared between Ctrl and SD mice 7 d after treatment with MTX or PBS (n = 7 for Ctrl and SD, n = 17 for Ctrl MTX, n = 11 for SD MTX). Results from two independent experiments were combined and are shown. A two-tailed Mann–Whitney U test was used for statistical analyses (P values are shown between the indicated groups). l, Plasma samples were collected from Ctrl and SD mice 2 d after MTX to detect soluble factors with the L308 antibody array (n = 5 for Ctrl MTX and SD MTX). Significant changes are summarized in a heat map. All dot plots and growth curves (at indicated time points) are shown as mean ± s.e.m. *P < 0.05, **P < 0.01, ***P < 0.001.
NATURE MEDICINE | VOL 25 | SEPTEMBER 2019 | 1428–1441 | www.nature.com/naturemedicine 1431 ARTICLES NATURE MEDICINE unable to reduce corticosterone levels in SD mice (Extended Data of corticosteroid medication, such as dexamethasone (DEX) and Fig. 4a). In a tumor-vaccination setting, increased corticosterone prednisolone (PRED), largely attenuated the efficacy of a pro- concentrations correlated with accelerated tumor growth (Extended phylactic cancer vaccine against tumor-cell re-challenge in non- Data Fig. 4b). MIFE injections reverted SD-mediated inhibition of stressed mice, and the therapeutic effect of MTX against MCA205 IFN-γ production by tumor-infiltrating T cells after chemotherapy fibrosarcomas in C57BL/6 mice and doxorubicin (DX) against or PD-1 blockade, and elevated circulating IFN-γ level in SD mice CT26 colorectal cancers in BALB/c mice (Fig. 3m,n and Extended treated with MTX (Fig. 3k,l). Conversely, systemic supplementation Data Fig. 4c–f). Moreover, DEX and PRED had no effects on tumor
a b 300 300 SD PBS ) 300 Ctrl PBS 300 SD PBS Ctrl PBS 2 )
Ctrl MTX SD MTX 2 Ctrl MTX SD MTX * ** = 0.4415 ns = 0.4643 ns 200 200 200 200 P P = 0.0159 P = 0.0079 P
100 100 100 100 TC-1 tumor size (mm
MCA205 tumor size (mm 0 0 0 0 0 10 20 30 0 10 20 30 0 5 10 15 20 25 0 5 10 15 20 25 Time (d) Time (d) Time (d) Time (d)
c dfe 350 Ctrl PBS Ctrl ISO Ctrl PBS Ctrl PBS ) 350 250 ) ) 400 2 2 SD PBS 2 SD ISO AR PBS AR PBS ) 300 300 2 Ctrl VAC Ctrl anti-PD-1 200 Ctrl MTX Ctrl MTX 250 SD VAC 250 SD anti-PD-1 AR MTX ** 300 AR MTX ** 200 1/7 TF 200 150 = 0.0087
P 200 150 150 2/8 TF 100 = 0.0022 * 100 100 P * 50 100 50 = 0.0133
50 P = 0.0233
3/6 TF TC-1 tumor size (mm P MCA205 tumor size (mm MCA205 tumor size (mm MCA205 tumor size (mm 9/10 TF 0 0 0 0 0 10 20 30 0 10 20 30 0 5 10 15 20 25 0 10 20 30 Time (d) Time (d) Time (d) Time (d)
g Ctrl MTX h ** ** ns ns * 40 2,000 P = 0.0022 10,000 P = 0.0022 300 20,000 1,500 P = 0.0418 SD MTX *** ** P = 0.1797 P = 0.2381 P = 0.0001 P = 0.0032 * 8,000 30 P = 0.0213 1,500 15,000 200 1,000 TILs (%)
+ ns 6,000 20 P = 0.3403 1,000 10,000 4,000 100 500 relative expression relative expression relative expression relative expression 10 500 5,000 relative expression 2,000 Ifng Cells in CD45 Rorc Gata3 Tbx21 0 0 0 0 Foxp3 0 0
DC T cell
Neutrophil Ctrl MTXSD MTX Ctrl MTXSD MTX Ctrl MTXSD MTX Ctrl MTXSD MTX Ctrl MTXSD MTX Macrophage
i j Ctrl SD
250 * P = 0.0399 10 *** P < 0.0001 *** P = 0.0005 ** P = 0.0052 *** P < 0.0001 ** P = 0.001 200 8 PBS
150 6 T cells (%) + γ 100 4 IFN- + spot counts in ELISPOT
γ 50 2 CD3 OXA IFN- 0 0
Ctrl SD Ctrl SD CD3 IFN-γ Hoechst 30 µm Ctrl SD Ctrl OXASD OXA Ctrl MTX SD MTX Ctrl MTX SD MTX
klAntibody array 400 ns P = 0.3623
) ** P = 0.0025 ** P = 0.0047 –1 2 log
300 2 1 (fold change) Ctrl MTX 200 0 −1 in serum (pg ml
γ 100 −2 SD MTX IFN- 0 LEP TEK IL15 IL25 IL22 IGF2 SELL XCL1 CSF3 CD80 CCL8 IL23A CCR6 NRTN
SD NRG3 EDAR EGFR IL3RA MFRP IFNB1
Ctrl ITGB2 INHBA FGF21 THBS1 CXCL9 CXCL1 IL1RL2 CCR10 IL31RA CXCR2 CXCR3 MFGE8 VCAM1 BMPER IGFBP7 TNFSF8 CXCL16 CXCL10 CD40LG SELPLG RETNLB CTNNB1 FCGR2B
SD MTX TNFSF10
Ctrl MTX ANGPTL2 TMPRSS5
1432 NATURE MEDICINE | VOL 25 | SEPTEMBER 2019 | 1428–1441 | www.nature.com/naturemedicine NATURE MEDICINE ARTICLES growth on their own. These results support the idea that glucocor- E, epithelial-associated (Itgae; also known as CD103) and zinc ticoids can subvert the therapeutic (and immunogenic) activity of finger and BTB domain containing 46 (Zbtb46) were not affected chemotherapy in vivo. It is noteworthy that relatively low doses of (Extended Data Fig. 5b). Reportedly, DEX can induce the expres- DEX failed to cause thymolysis (Extended Data Fig. 4g, h), but were sion of immunosuppressive genes, including Tnfaip3, nuclear able to suppress MTX-induced recruitment of DC and neutrophils, factor of kappa light polypeptide gene enhancer in B cells inhibi- and IFN-γ production by CD3+ T lymphocytes in the tumor micro- tor, alpha (Nfkbia) and dual specificity phosphatase 1 (Dusp1) environment (Extended Data Fig. 4i–l). (refs. 24,25). However, SD did not affect Tnfaip3 expression and Altogether, these results support the hypothesis that SD-induced caused a clear reduction in the expression of Nfkbia and Dusp1 elevation in the glucocorticoid tonus can occur independently of (Extended Data Fig. 5c). SD also triggered noticeable transcrip- the serotoninergic and adrenergic systems and induce a state of tional changes in some Toll-like receptors (TLRs), chemokines, immunosuppression that renders chemotherapy and immunother- cytokines and receptors and immune checkpoint molecules in apy inefficient. TIDCs (Extended Data Fig. 5d–g). More importantly, SD signif- icantly augmented the level of messenger RNA encoding gluco- Social-defeat-induced upregulation of Tsc22d3 in tumor-infil- corticoid-inducible transcriptional regulator Tsc22 domain family trating dendritic cells. Tumor antigen processing and presentation protein 3 (Tsc22d3; also known as glucocorticoid-induced leucine 7,9 are prerequisites for optimal antitumor TH1 and/or TC1 responses . zipper, Gilz) (Fig. 4g), which functions as a potent mediator of Of note, SD affected the expression of genes involved in major his- anti-inflammation and immunosuppression26–28. In addition, SD tocompatibility complex (MHC) I- and MHC II-restricted antigen- modulated the transcription of a large set of Tsc22d3 downstream presentation pathways in TIDCs (Fig. 4a,b). Also, SD pretreatment target genes identified by ingenuity pathway analysis (Fig. 4g). We reduced the proliferation of tumor antigen OVA-specific, CD45.1+ performed a time-course analysis of Tsc22d3 expression in four OT1 cells in tumor-draining lymph nodes after MTX-based che- major immune cell subsets (DCs, macrophages, neutrophils and motherapy, as indicated by reduced dilution of the cell-prolifera- T cells) purified from tumors, spleens and tumor-draining lymph tion dye eFluor670 in these cells (Fig. 4c). Although SD did not nodes from both control and SD mice after chemotherapy. Of alter the quantity of TIDCs 2 d after chemotherapy with MTX note, in response to social defeat, Tsc22d3 was remarkably upregu- (Fig. 2g), it did modify the ‘quality’ of TIDCs, as measured by mas- lated only in TIDC on day 2 and day 7 post-MTX. No other cell sive changes in their transcriptomes (Fig. 4d). RNA sequencing type, either in tumors or lymphoid organs, exhibited a significant (RNA-seq) analyses revealed signs of SD-induced local immune increase in Tsc22d3 mRNA level in response to stress (Fig. 4h). The suppression, as indicated by the downregulation of genes involved SD-triggered elevation in Tsc22d3 mRNA in TIDCs relied on glu- in the processes of myeloid leukocyte differentiation, antigen sig- cocorticoid receptor signaling because repeated MIFE injections naling pathway, neutrophil chemotaxis and immune-system pro- abrogated Tsc22d3 upregulation in TIDCs of SD mice (Fig. 4i). cesses. Gene ontology (GO) analysis suggested that SD amplified Serum from SD or AR mice stimulated Tsc22d3 mRNA expression the circadian rhythm and cellular metabolism networks in TIDCs by bone-marrow-derived dendritic cells (BMDCs) in vitro, which (Fig. 4e,f and Extended Data Fig. 5a). SD-induced stress did not correlated with corticosterone elevation (Fig. 4j and Extended affect the lineage commitment and differentiation of DCs as the Data Fig. 5h). This Tsc22d3-mRNA-inducing effect of ‘stressed’ expression of C-type lectin domain family 9, member a (Clec9a; sera on BMDCs was lost following addition of MIFE (Fig. 4k and encoding a protein also known as DNGR-1), integrin alpha Extended Data Fig. 5i). Tsc22d3 mRNA was not upregulated in
Fig. 3 | The secretion of stress-related hormones and neurotransmitters and their impact on anticancer therapies. a,b, Plasma corticosterone concentrations in Ctrl (n = 17) and SD (n = 18) mice were detected by enzyme-linked immunosorbent assay (ELISA) at day 10 of 10-d SD. An unpaired two-tailed Student’s t-test was used for statistical analyses (a). Linear regression analysis was used to explore the correlation between corticosterone levels and total locomotion from light–dark box tests for individual mice. Goodness of fit was quantified and is shown as r2 (b). A representative result from three independent experiments is shown. c,d, Plasma serotonin levels were evaluated with ELISA in both Ctrl (n = 13) and SD (n = 15) mice at day 10. An unpaired two-tailed Student’s t-test was used for statistical analyses (c). The correlation between serotonin levels and the total locomotion of each mouse was analyzed with linear regression analysis (d). A representative result from two independent experiments is shown. e,f, ELISA-based detection of plasma norepinephrine levels (e) and their inter-relationship with total locomotion in each mouse of Ctrl (n = 15) and SD (n = 10) groups (f). An unpaired two-tailed t-test (e) and linear regression analysis (f) were used for statistical analyses. A representative result from two independent experiments is shown. g, The levels of ACTH (left) and corticosterone (right) in Ctrl and SD mice were quantified with ELISA at different time points. On day –12, –7, –2, 7 and 12: n = 5, 5, 13, 10 and 10 for Ctrl mice; n = 7, 7, 11, 9 and 9 for SD mice. A two-tailed Mann–Whitney U test was used for statistical analyses (P values are shown between indicated groups). h,i, Impact of MIFE on tumor growth in Ctrl and SD mice treated with PBS (left) versus MTX (right, n = 5 for all groups) (h), or isotype control (ISO) versus anti-PD-1 antibodies (i) (n = 6 for Ctrl PD-1 and SD PD-1, n = 7 for Ctrl ISO, n = 5 for all the other groups). A representative result from two independent experiments is shown. Two-tailed Mann–Whitney U test was used for statistical analyses of the area under the curve (P values are shown between indicated groups). j, Effect of MIFE on the efficacy of anticancer vaccination in Ctrl (left) and SD (right) mice (n = 6 for Ctrl, Ctrl VAC and Ctrl VAC MIFE, n = 5 for all the other groups). A representative result from two independent experiments is shown. A two- tailed Mann–Whitney U test was used for statistical analyses of the area under the curve (P values are shown between indicated groups). k, Impact of MIFE on IFN-γ production by tumor-infiltrating T cells from Ctrl and SD mice treated with PBS, MTX or anti-PD-1 as measured by ELISPOT (n = 8 for Ctrl and SD, n = 13 for Ctrl MIFE and SD MTX MIFE, n = 14 for Ctrl MTX MIFE, n = 12 for SD MIFE, Ctrl anti-PD-1 and SD anti-PD-1, n = 10 for Ctrl anti-PD-1 MIFE, n = 11 for SD anti-PD-1 MIFE). Results from two independent experiments were combined and are shown. A two-tailed MannWhitney U test was used for statistical analyses (P values are shown between indicated groups). l, Effect of MIFE on plasma IFN-γ levels from MCA205-bearing Ctrl or SD mice treated with MTX, as quantified by ELISA (n = 9 for Ctrl MTX, n = 8 for SD MTX, n = 7 for Ctrl MTX MIFE, n = 6 for SD MTX MIFE). m, Impact of DEX on the efficacy of the prophylactic tumor vaccine (n = 5 for PBS and PBS DEX, n = 6 for VAC, n = 7 for VAC DEX). A representative result from two independent experiments is shown. A two-tailed Mann–Whitney U test was used for statistical analyses of the area under the curve (P values are shown between indicated groups). n, Tumor-bearing mice were treated with MTX against MCA205, in the presence of DEX or solvent control. Tumor progression was compared between the indicated groups. A two-tailed MannWhitney U test was used for statistical analyses of the area under the curve (n = 4 for MTX DEX 20, n = 5 for all the other groups). A representative result from two independent experiments is shown. All dot plots and growth curves are shown as mean ± s.e.m. *P < 0.05, **P < 0.01, ***P < 0.001.
NATURE MEDICINE | VOL 25 | SEPTEMBER 2019 | 1428–1441 | www.nature.com/naturemedicine 1433 ARTICLES NATURE MEDICINE
BMDCs after mixed culture with live or MTX-treated cancer cells, T cells, nor in splenic T cells exposed to ‘stressed’ sera from AR yet was following addition of DEX (Extended Data Fig. 5j). Tsc22d3 or SD mice ex vivo (Extended Data Fig. 5k,l). The abundance of expression was neither augmented by SD in tumor-infiltrating mRNA encoding glucocorticoid receptor (Nr3c1) appeared to be
ab cd e *** P < 0.0001 20 * Ctrl * P = 0.0496 400 7,500 Ctrl 7,500 ) 20
) P = 0.0102
SD –1 –1 SD )
6,000 –1 6,000 300 15 15 g ml
4,500 µ 4,500 ( 200 10 10 3,000 3,000 100 5 2 5 1,500 2 1,500 r = 0.254 r = 0.361 Serotonin Total locomotion (cm) Total locomotion (cm) P = 0.0063 Norepinephrine (ng ml Corticosterone (ng ml P < 0.0001 0 0 0 0 0 Ctrl SD 0 100 200 300 400 Ctrl SD 0 5 10 15 20 Ctrl SD –1 Corticosterone (ng ml–1) Serotonin (µg ml )
f g Ctrl SD Ctrl SD 9,000 Ctrl 1.5 900 ns ns ns ) ns SD * ** *** *** * P > 0.999 ** P = 0.0456 P = 0.8208 –1 P = 0.966 P = 0.0025 P = 0.2389 P = 0.0051 P < 0.0001 P < 0.0001 P = 0.0103 r 2 = 0.009 ) –1 6,000 P = 0.6580 1.0 600
3,000 0.5 300 ACTH (ng ml Total locomotion (cm) Corticostrone (ng ml 0 0 0 0 5 10 15 20 Day –12 –7 –2 7 12 Day –12 –7 –2 7 12 Norepinephrine (ng ml–1) 10-d SD Tumor inoculation 10-d SD Tumor inoculation (day 0) (day 0)
hiCtrl PBS Ctrl MTX Ctrl ISO SD ISO SD PBS SD MTX Ctrl ISO MIFE SD ISO MIFE SD MTX MIFE
) SD PBS MIFE
) SD anti-PD-1
2 300 300 Ctrl anti-PD-1 450 2 450 Ctrl anti-PD-1 MIFE SD anti-PD-1 MIFE 250 250 * = 0.4415 ns **
200 200 ** 300 P 300
150 150 ** = 0.0238 = 0.0079 P = 0.0023 P
100 100 150 P 150 = 0.0043 50 50 P MCA205 tumor size (mm 0 0 MCA205 tumor size (mm 0 0 0 10 20 30 0 10 20 30 0 10 20 30 0 10 20 30 Time (d) Time (d) Time (d) Time (d)
jk Ctrl Ctrl MIFE SD SD MIFE ** P = 0.0029 * P = 0.0106 300 500 ) Ctrl VAC 5/6 TF SD VAC 1/5 TF P = 0.0003 ns P = 0.395 ** 2 350 350 *** P = 0.0036 Ctrl VAC MIFE SD VAC MIFE 3/5 TF 300 2/6 TF 300 400 * P = 0.0372 250 250 200 300 200 200 ** 150 ** 150 200 * spot counts in ELISPOT
γ 100 100 100 = 0.0022 = 0.0043 100 P P 50 50 IFN- = 0.0435 P
MCA205 tumor size (mm 0 0 0 0 0 10 20 30 0 10 20 30 Ctrl SD Time (d) Time (d) Ctrl MIFESD MIFE Ctrl anti-PD-1SD anti-PD-1 Ctrl MTXSD MIFE MTX MIFE lnm PBS PBS DEX Ctrl anit-PD-1SD anit-PD-1 MIFE MIFE VAC VAC DEX 400 * ) ) P = 0.036 100 PBS 400 MTX 6/6 TF 2 400 –1 ** P = 0.0047 DEX 100 MTX DEX 100 *
300 ) 80 DEX 20 MTX DEX 20 300 300
) DEX 5 MTX DEX 5 ** *** –1 60 = 0.6667 ns = 0.0149 P 200 P 200 200 40 3/7 TF (ng ml = 0.0079 = 0.0007 P P in serum (pg ml
γ 100 Tumor free (% 100 100 20 0/5 TF
IFN- 0/5 TF
0 0 MCA205 tumor size (mm 0 0 0 10 20 30 40 50 0 10 20 30 0 10 20 30 Time (d ) Time (d ) Time (d ) Ctrl MTXSD MTX
Ctrl MTXSD MIFE MTX MIFE
1434 NATURE MEDICINE | VOL 25 | SEPTEMBER 2019 | 1428–1441 | www.nature.com/naturemedicine NATURE MEDICINE ARTICLES irrelevant to the preferential induction of Tsc22d3 expression in beta 1 (Ifnb1) and C-C motif chemokine ligand 5 (Ccl5)) remained TIDCs because the level of Nr3c1 transcription was comparable largely unchanged in TIDCs lacking Tsc22d3 compared with their between DCs, macrophages, neutrophils and T cells residing in the WT counterparts (Extended Data Fig. 6g–i). The genetic manipu- tumor microenvironment (Extended Data Fig. 5m). lation of Tsc22d3 levels in DCs (in both Tsc22d3-Tg and Tsc22d3- Altogether, the transcriptomic analyses support the notion cKO mice) did not affect the frequency of DC, macrophages or that Tsc22d3 is specifically upregulated in TIDCs in response to neutrophils in tumors 2 d after MTX. Tsc22d3-cKO mice exhib- stress conditions. ited a significant increase in the density of tumor-infiltrating T lymphocytes as compared with WT controls (Extended Data Glucocorticoid-induced Tsc22d3 mediates immunosuppression. Fig. 6j,k). Also, enzyme-linked immune absorbent spot (ELISPOT) Encouraged by the above findings, we decided to investigate assays performed 7 d post-MTX or post-PD-1 blockade indicated whether TSC22D3 in the DC compartment has functional rel- that the DC-specific Tsc22d3 knockout enhanced IFN-γ secre- evance in cancer therapies. Without SD preconditioning, trans- tion by tumor-infiltrating T cells under SD conditions (Fig. 5j). gene-enforced overexpression of Tsc22d3 in DCs (under the Consistently, tumor antigen OVA-specific T cell proliferation was control of the integrin alpha X (Itgax) promoter, abbreviated as inhibited under SD conditions in WT mice, but not in Tsc22d3- Tsc22d3-Tg)27 abolished the efficacy of immunogenic chemother- cKO mice (Extended Data Fig. 6l). apy DX and suppressed the cancer-preventive immunity elicited As compared to age- and sex-matched healthy volunteers, by a prophylactic vaccination (Fig. 5a,b and Extended Data Fig. patients with colorectal cancer or NSCLC exhibited an increase 6a). Conversely, conditional knockout (cKO) of Tsc22d3 in DCs in circulating cortisol levels (Fig. 6a), a finding that was validated (achieved by a Cre recombinase expressed under the control of the in an independent cohort (Extended Data Fig. 7a,b). In addition, Itgax promoter that excises the last exon of Tsc22d3 flanked by two peripheral blood mononuclear cells (PBMCs) from such patients loxP sites, abbreviated as Tsc22d3-cKO)28,29 was sufficient to reverse exhibited increased TSC22D3 transcription compared with healthy the DEX-induced impairment of the efficacy of immunogenic che- volunteers (Fig. 6b). Sera from patients with cancer (but not those motherapy in vivo (Fig. 5c and Extended Data Fig. 6b). Moreover, from healthy donors) were able to boost TSC22D3 transcription DC-targeted deletion of Tsc22d3 improved therapeutic responses by PBMCs from healthy volunteers, which could be reversed by to both PD-1 blockade and the immunogenic drug MTX in the adding MIFE (Fig. 6c). Notably, elevated cortisol concentrations context of SD (Fig. 5d,e and Extended Data Fig. 6c,d). Similar were associated with negative mood in patients with cancer, as to their WT counterparts, Tsc22d3-cKO mice showed a normal reflected by the profile of mood states (POMS) questionnaire32,33 response to SD with respect to the anxiety-related behavior and (Fig. 6d). We also found a positive correlation between plasma cor- the elevation of circulating corticosterone levels (Extended Data tisol levels and TSC22D3 expression by PBMCs in patients with Fig. 6e,f). Next, we investigated the possibility that SD-induced cancer (Fig. 6e). Hence, elevated endogenous glucocorticoid level immunosuppression might require Tsc22d3 expression by DCs. is linked to mental stress as well as to increased TSC22D3 expres- In this context (SD and MTX-based chemotherapy of established sion by PBMCs. No obvious increase in ACTH levels was observed tumors), TIDCs were purified from WT and Tsc22d3-cKO mice, in the same cohorts of patients compared with healthy volunteers. followed by RNA-seq-based transcriptome analyses (Fig. 5f,g). The Also, there was no correlation between ACTH and cortisol levels absence of Tsc22d3 favored the upregulation of IFN-β-stimulated (Extended Data Fig. 7c,d). genes in TIDCs (Fig. 5h), indicating that Tsc22d3 represses the In multiple human tumor types, TSC22D3 mRNA expression type I IFN response, which is known to be required for anticancer can be detected in tumor-infiltrating immune cells, including DCs immunosurveillance30,31. Interestingly, the SD-inhibited expression (Extended Data Fig. 7e). TSC22D3 transcription positively corre- of several genes involved in MHC I-restricted antigen presenta- lated with the presence of macrophages34, especially in cancers of tion was partially reversed (calreticulin (Calr), Sec61 beta subunit the digestive tract (for example esophagus, stomach, colon, rectum (Sec61b), Sec61g) in Tsc22d3-deficient TIDCs (Fig. 5i). However, and pancreas), lung and melanoma. For other cancer types, and the expression of TLRs, antigen-uptake and cell-adhesion mol- other major immune-cell populations, the correlation was less con- ecules, chemokines, cytokines and receptors (except interferon sistent (Extended Data Fig. 7f). Remarkably, single-cell sequencing
Fig. 4 | Mental stress modulates the function and transcriptome of tumor-infiltrating dendritic cells (TIDC). a,b, TIDCs were purified from Ctrl or SD mice 2 d after MTX. The expression of genes involved in the MHC I- or MHC II-restricted antigen presentation pathway was tested by quantitative RT–PCR (qRT–PCR) (n = 8 for Ctrl MTX, n = 7 for SD MTX). A representative result from two independent experiments is shown. A two-tailed Mann– Whitney U test was used for statistical analyses (P values are shown between the indicated groups). c, Proliferation of tumor antigen OVA-specific CD45.1+OT1 cells in the draining lymph nodes of Ctrl and SD mice after MTX, as determined by eFlour 670 dye dilution (n = 6 for Ctrl MTX, n = 5 for SD MTX). Representative fluorescence histograms (left) and mean fluorescence intensities (MFI) (right) of OT1 cells from two independent experiments are shown. A two-tailed Mann–Whitney U test was used for statistical analyses. d–g, Heat map revealing SD-induced changes in the gene-expression profile of TIDCs (d). Gene ontology analysis of up- and downregulated genes. Gene ratio (shown in green) and –log10(P) of all GO terms are shown (e). GO network analysis of significantly downregulated genes in TIDCs from SD mice, as compared with TIDCs from Ctrl mice (f). Heat map showing the expression of Tsc22d3 and its target genes in TIDC from Ctrl and SD mice (g). n = 3 for both Ctrl MTX and SD MTX. Color bars in d,g indicate normalized expression. h, Heat map summary of the fold changes to Tsc22d3 mRNA expression in SD compared with Ctrl mice, in different immune-cell subsets purified from tumors or lymphoid organs at different time points after MTX-based chemotherapy. Expression was determined by qRT–PCR using Ppia as an internal control (on day 2, 7 and 14: n = 8, 12 and 5 for both Ctrl MTX and SD MTX groups). Color bar indicates fold change. i, The impact of repeated MIFE injections on Tsc22d3 expression in TIDC from Ctrl or SD mice, treated with either PBS or MTX (n = 6 for Ctrl MTX and Ctrl MTX MIFE, n = 7 for SD MTX, n = 5 for SD MTX MIFE). A representative result from two independent experiments is shown. A two-tailed Mann–Whitney U test was used for statistical analyses (P values are shown between the indicated groups). j, Serum from Ctrl and SD mice was collected and used to stimulate BMDCs. The correlation between serum corticosterone levels and Tsc22d3 transcription was analyzed with linear regression test (n = 4 for Ctrl serum, n = 12 for SD serum). k, Tsc22d3 expression in BMDCs treated with serum from individual Ctrl or SD mice, in the presence of glucocorticoid receptor antagonist MIFE or PBS solvent alone (n = 4 for Ctrl and Ctrl MIFE, n = 12 for SD and SD MIFE). A two-tailed Mann–Whitney U test was used for statistical analyses (P values are shown between the indicated groups). Data from qRT–PCR analyses are shown as mean ± s.d. All the other dot plots are shown as mean ± s.e.m. In heat maps, color scales represent the values of log2-transformed fold changes. *P < 0.05, **P < 0.01, ***P < 0.001.
NATURE MEDICINE | VOL 25 | SEPTEMBER 2019 | 1428–1441 | www.nature.com/naturemedicine 1435 ARTICLES NATURE MEDICINE
analyses of TIDCs and matched PBMCs from patients with lung can- analyses35,36 of overall survival (OS) suggested a negative prognostic cer revealed a positive correlation between TSC22D3 mRNA levels in value for high TSC22D3 expression in patients with lung cancer with PBMCs and TSC22D3 expression in TIDCs (Fig. 6f). Kaplan–Meier smoking experience, resembling the results obtained in the murine
ab5 MHC I pathway 4 MHC II pathway Ctrl MTX Ctrl MTX * ** * SD MTX SD MTX P = 0.0295 P = 0.0041 P = 0.0428 4 ** ** * ** 3 P = 0.0093 P = 0.0039 P = 0.0483 P = 0.0093 ** 3 P = 0.0093 2 2 fold change fold change 1 Relative expression
Relative expression 1
0 0
Ifi30 Calr Tap1 Tap2 Ctsb Ctsh Ctso Ctss Ctsz Canx Erap1 Pdia3 Lamp1 Sec61b Sec61g
c deCtrl MTX SD MTX DOWN UP Ctrl 2 15 SD 1 2,500 * P = 0.0303 12 Gene
0 9.3% 8.3% 9.3% 2.7%
4.7% ratio 7.3% 3.3% −1 6.0% 8.7% 6.7% 2,000 −2 9 value) P 9.8% (
7.1% 1,500 10 6
–log 3 1,000 0
MFI eFluor 670 of OT1 500
production γ 0 IFN- T cell activation CircadianRhythmic rhythm process
Adaptive immune response 0 1 2 3 4 5 6 Heterotypic cell–cell adhesion Ctrl MTX SD MTX RegulationRegulation of defense of cell–cell response adhesion 10 10 10 10 10 10 10 eFlour 67 0 PositiveRegulation regulation of of immune cell–cell effector adhesion process Regulation of leukocyte cell–cell adhesion Regulation of heterotypic cell–cell adhesion f g Endogenous stimulus response hormone Ctrl MTXSD MTX Electron transport chain Cell differentiation involved proliferation Klf9 Neutrophil chemotaxis 1.5 Gpr34 Inflammatory IL-6 production Signal transmission event Tsc22d3 response stress 1 intracellular Anion Ube3a Biotic homeostasis Rasa1 unfolded 0.5 Cxcl9 Blood coagulation hemostasis Tlr8 Peptidyl-tyrosine 0 Fkbp5 phosphorylation modification Tlr4 Insulin bone resorption Lymphocyte-mediated cytotoxicity −0.5 Qser1 Tlr3 Activation regulation Fmr1 Antigen signaling pathway −1 cell lymphocyte Cep55 response-activating −1.5 Nufip2 Celluar ion homeostasis Tlr7 Response based immune Spongiotrophoblast differentiation Ddx6 immunogobulin Rictor embryonic placenta Bbc3 Mitochondrial membrane Sirt6 Primary metabolic process + + Ccl5 CD4 CD25 regulatory Aldh1l1 Cxcl3 Aldh3b1 Myeloid leukocyte differentiation Carbohydrate alcohol Dusp1 Ccl2 Immune system process Csf2 Il10 Thyroid hormone transport Regulation MAPKKK cascade Mif Nfkbib Induction programmed Ccl3 Regulation protein death apoptosis Col1a1 Triglyceride mobilization lipid localization secretion Pck1 Fasl Localization Complex component Cxcl2 establishment assembly Cxcl1 Process regulation activity negative Ifng
hiTsc22d3 expression TIDC j BMDC k BMDC ** P = 0.005 1 4 fold change (SD MTX vs Ctrl MTX) 8 5 *** P < 0.0001 * Ctrl serum P = 0.0221 3 P = **0.006 P = 0.02* 2 DC SD serum ns P = 0.6132 r 4 Macrophage 6 3 * P = 0.0454 2 Neutrophil 3
3 ** 3 Tumo 3 P = 0.0017 1 T cell 4 2 0 DC Tsc22d 2 Tsc22d Tsc22d P = 0.04* 2 P = 0.04* 6 Macrophage −1 2 1 2
Neutrophil relative expression relative expression relative expression Spleen r = 0.3336 1 −2 T cell P = 0.0191 DC 0 0 0 −3 LN P = 0.04* 5 T cell 0 500 1,000 1,500 Ctrl SD Day 2 7 14 Corticosterone (ng/ml) After chemotherapy Ctrl MTXSD MTX Ctrl MIFESD MIFE Ctrl MTXSD MIFE MTX MIFE
1436 NATURE MEDICINE | VOL 25 | SEPTEMBER 2019 | 1428–1441 | www.nature.com/naturemedicine NATURE MEDICINE ARTICLES
NSCLC model induced by a tobacco constituent urethane (Fig. 6g). immunosuppressive effects through its ability to homo- and het- For patients bearing digestive cancers, such as gastric and colorec- erodimerize with partner proteins, such as nuclear factor kappa- tal carcinomas, high expression of TSC22D3 also correlated with light-chain-enhancer of activated B cells (NF-κB), RAS, and unfavorable OS (Fig. 6h,i). In accordance with the fact that TSC22D3 CCAAT-enhancer-binding protein (C/EBP)26. TSC22D3 also binds represses type I IFN responses in mice, low expression of C-X-C to SMAD2, leading to SMAD2 phosphorylation and consequent 39 (CXCL9) and CCL5 correlated with unfavorable OS in patients with optimal induction of FOXP3 and Treg proliferation . Previous stud- lung cancer who had a history of smoking, as well as in patients with ies have demonstrated that TSC22D3 expression by DCs is suffi- colorectal or gastric cancer (Extended Data Fig. 7g,h). cient to confer on them a tolerogenic phenotype40,41. Hence, it is We conclude that TSC22D3 plays a major role in stress- or tempting to assume that TSC22D3 constitutes a ‘checkpoint’ for an glucocorticoid-induced cancer-therapy-relevant immunosuppres- anticancer immune response. Hence, TSC22D3 should be regarded sion. When highly expressed in DCs, TSC22D3 suffices to subvert as an actionable target that, if removed or neutralized, might favor cancer-therapy-induced tumor-growth inhibition, while its removal the re-establishment of immunosurveillance against cancer. The from DCs abolishes glucocorticoid-mediated immunosuppression, absence of Tsc22d3 in TIDCs potently augmented the type I IFN including in the context of stress. These findings urge caution in response, consistent with an earlier report showing that lipopoly- stress management of patients and applications of synthetic gluco- saccharide-challenge-induced serum IFN-α was significantly corticoids in immune-dependent cancer therapies. higher in GRCD11c-cre mice (which lack the glucocorticoid receptor in DCs) than that in WT mice42. Moreover, it has been reported Discussion that polyinosinic:polycytidylic acid can trigger more IFN-α pro- Here, we provide evidence that endogenous glucocorticoids that duction in monocytes from glucocorticoid-resistant patients with increase in response to repeated social defeat (SD) or acute restraint AIDS compared with that from glucocorticoid-sensitive patients (AR) stress may mediate potent immunosuppressive effects, link- with AIDS and controls43. Thus, the glucocorticoid receptor and ing stress with cancer-therapy failure. Indeed, both stress (which its downstream effector molecule TSC22D3 act in a similar way to increases the endogenous levels of circulating corticosterone in restrain the type I IFN response. That said, the mechanisms linking mice and cortisol in humans) and the exogenous supply of glu- subtle elevations in glucocorticoid tonus to a TIDC-specific upreg- cocorticoids are sufficient to subvert the tumor-growth-reducing ulation of TSC22D3 (which might be linked to a cell-type-specific effects of immunogenic chemotherapy with OXA or anthracyclines. epigenome organization44) remain to be elucidated. Conversely, MIFE, an antagonist of the glucocorticoid receptor, Cancer chemotherapies are often accompanied by ‘co-medica- restores the efficacy of immunogenic chemotherapy in mice that tions’ that include glucocorticoids, used for instance for suppress- had been exposed to SD, providing evidence in favor of the immu- ing mucositis, nausea, vomiting, fever and other side effects45. Our nosuppressive impact of endogenous glucocorticoids10. Previous results suggest that exogenous glucocorticoids might interfere with works dealing with the effects of stress on cancer have discovered the treatment-relevant immunostimulatory action of chemother- the capacity of another stress hormone, epinephrine, to stimulate apy, a suspicion that requires thorough retrospective or prospec- tumor vascularization37 and cancer stem-cell-like properties38. tive evaluation in patients with cancer. Beyond this effect, one may Thus, several mechanisms may co-occur to connect mental stress to consider strategies to reduce anxiety and depression in oncologi- oncogenesis and tumor progression. However, at least in the context cal patients with the scope of reducing endogenous glucocorticoids of immune-dependent anticancer therapies, glucocorticoids play a and their immunosuppressive effects46,47. Such studies would benefit more prominent role than the adrenergic system. from POMS assessment, systematic measurement of endogenous Interestingly, a glucocorticoid-inducible molecule, TSC22D3 glucocorticoids and TSC22D3 expression analysis in an attempt to (GILZ) plays an important role in stress-induced immunosup- introduce an objective (and mechanistic) link between psychoso- pression. Thus, SD preconditioning augments the transcription cial or psychopharmacological interventions and the outcome of of Tsc22d3 mRNA in the DC compartment of mice. Moreover, oncological treatments. It remains an open speculation that inter- transgenic overexpression of Tsc22d3 within DCs was sufficient ference of hypothalamic–pituitary–adrenal circuitries accounting to abolish tumor growth inhibition by chemotherapy and tumor for the increase in endogenous glucocorticoids or direct blockade vaccine in unstressed mice, whereas DC-targeted knockout of of the glucocorticoid receptor (with MIFE, which is Food & Drug Tsc22d3 abrogated the immunosuppressive effects of endogenous Administration (FDA)-approved for the treatment of hypercorti- or exogenous glucocorticoids with respect to the outcome of can- solism in adults with Cushing’s syndrome)48 might help improving cer therapies. TSC22D3 is a transcriptional regulator that mediates the efficacy of chemotherapy and immunotherapy.
Fig. 5 | DC-targeted genetic manipulation of Tsc22d3 modulates the efficacy of anticancer therapies and the transcriptome of TIDCs. a,b, In non- stressed conditions, the responsiveness of established tumors growing in Tsc22d3-Tg mice (overexpressing Tsc22d3 in DC, n = 9 for both groups) and litter-mate controls (n = 7 for PBS, n = 5 for DX) to DX were compared. A representative result from two independent experiments is shown. A two-tailed Mann–Whitney U test was used for statistical analyses (a). Moreover, the response to prophylactic tumor vaccine (VAC) was compared. For litter-mate controls, n = 11 for PBS, n = 18 for VAC. For Tsc22d3-Tg mice, n = 7 for PBS, n = 8 for VAC. A representative result from two independent experiments is shown. A log-rank test was used for statistical analyses (b). c, Upon treatment with DEX, the efficacy of MTX against MCA205 tumors was compared between WT control and Tsc22d3-cKO mice. A representative result from two independent experiments is shown. A two-tailed Mann–Whitney U test was used for statistical analyses. d,e, The therapeutic effects of PD-1 blockade (d) or MTX-based chemotherapy (e) were determined in WT and Tsc22d3-cKO mice preconditioned with SD. A representative result from two independent experiments is shown. A two-tailed Mann–Whitney U test was used for statistical analyses. f, Differential gene expression in TIDCs from WT versus Tsc22d3-cKO mice that received SD and MTX-based chemotherapy was determined by RNA-seq. Significantly up- and downregulated GO terms are summarized (n = 3 for WT, n = 5 for Tsc22d3 cKO). g, Heat map showing genes that were inhibited in WT mice following SD (n = 3) versus Ctrl (n = 3), and were upregulated in Tsc22d3-cKO mice (n = 5) versus WT mice (n = 3) under the SD condition. h,i, Type I IFN response (h) and MHC I- and MHC II-restricted antigen presentation pathway (i) gene signatures were analyzed by RNA-seq. j, ELISPOT-based analyses of IFN-γ secretion by tumor-infiltrating T lymphocytes in Ctrl and SD mice that received anti-PD-1 or MTX. Results from two independent experiments were combined and are shown; n values are indicated on the graph. A two-tailed Mann–Whitney U test was used for statistical analyses. All dot plots and growth curves are shown as means ± s.e.m. In heat maps, color scales represent the values of log2-transformed fold changes. *P < 0.05, **P < 0.01, ***P < 0.001.
NATURE MEDICINE | VOL 25 | SEPTEMBER 2019 | 1428–1441 | www.nature.com/naturemedicine 1437 ARTICLES NATURE MEDICINE
In summary, our results suggest it is possible to increase Online content anticancer immunosurveillance by neutralizing or intercept- Any methods, additional references, Nature Research reporting ing endogenous glucocorticoids and their downstream effec- summaries, source data, statements of code and data availability and tor TSC22D3. Both targets might constitute actionable immune associated accession codes are available at https://doi.org/10.1038/ checkpoints. s41591-019-0566-4.
a b 350 350 100 ) Littermate PBS Littermate PBS
2 Tsc22d3-Tg PBS 15/18 TF 300 300 Littermate VAC Littermate DX Tsc22d3-Tg DX 80 Tsc22d3-Tg PBS 250 250 **
Tsc22d3-Tg VAC
60 *** 200 200 = 0.3865 ns = 0.0051 P 150 P 150 40 < 0.0001 P
100 100 Tumor-free (%) 20 1/8 TF 50 50 0/11 TF 0/7 TF MCA205 tumor sizes (mm 0 0 0 0 5 10 15 20 25 0 5 10 15 20 25 0 10 20 30 40 Time (d) Time (d) Time (d)
WT PBS DEX (n = 4) cdWT SD ISO (n = 4) e WT SD PBS (n = 4) WT MTX DEX (n = 6) WT SD anti-PD-1 (n = 6) WT SD MTX (n = 5) Tsc22d3-cKO PBS DEX (n = 4) Tsc22d3-cKO SD ISO (n = 5) Tsc22d3-cKO SD PBS (n = 4) Tsc22d3-cKO MTX DEX 400 400 Tsc22d3-cKO SD anti-PD-1 400 Tsc22d3-cKO SD MTX (n = 5)
) (n = 5) ) (n = 6) ) 2 2 2 m m * 300 300 300 * = 0.1143 ns P = 0.0159 ** P
= 0.1111 ns 200 200 200 P = 0.0159 P **
= 0.0043 P 100 100 100 = 0.0087 P MCA205 tumor size (m MCA205 tumor size (m MCA205 tumor sizes (mm 0 0 0 0 5 10 15 20 25 0 5 10 15 20 25 0 5 10 15 20 25 Time (d) Time (d) Time (d)
f DOWN UP g 5 i 1 6 a 12 t 7 3 1 e 5 1 9 2 3 4 2 1 8 7 6 1 8 1 2 1 1 Gene 3 p 2.9%
ratio Klra Serpinb2 Gm1094 Skap Tmem40 Cxcl Gata Tnp2 Rpl3 Mt2 Crispld2 Zscan2 Rpl3 F3 Rpl3 Rps27r Taf9 Cdc42ep Gstp Hist1h2a Hist1h1e Vgll3 Ndufb Ly6c Ndufa Psmg4 Rnf144 Apol9b Mg Ankrd1 Bst2 Hoxc Ifitm3 Rps1 Tmsb10 Pcsk Rpl3 Fosl Isg20 Ddit3 Cd52 Gm1221 Ms4a4d Ccdc12 Irf7 Tceb Rpl2 S100a11 Flt1 Pfdn2 Sepw Cebp Frs3 Pttg Pfdn1 Tnfrsf12a H2 − T22 Ifnb1 Plekhn1 Rpl18a Rpl2 Ccl5 Fcer1g Irf1 ) 9 X 1.5 5.3% 2.2% 5.9%
6.2% 1 5.3% 5.8% 4.8%
2.8%
3.8%
2.8% 8.0% 2.2% 6.5% P value 4.8% 0.5
( 6
Ctrl MT 0 10 WT
X −0.5
–log 3 −1 −1.5 SD MT 0 s e a 2 WT
X 1 0 Lipid transport SD MT Sterol transport -cKO Lipid localization −1 Cholesterol efflux 3
Cholesterol transport −2 Leukocyte chemotaxis Tsc22d Epithelial cell proliferation Alcohol metabolic proces Regulation of epithelial cell Regulation of angiogenesis Response to interferon-bet Cellular response to interferon Regulation of vasculature dev Organic hydroxy compound trans Organic hydroxy compound metab
h i j ** P = 0.0046 Type I IFN responses MHC I and MHC II pathway 200 ns P = 0.1448 ** P = 0.0022 ** P = 0.002 X WT SD MTX Tsc22d3-cKO SD MTX ns P = 0.2045 ** P = 0.0041 150 Ifnb1* ns P = 0.3985 Gm4841* Gm4951* Xaf1* WT SD MT 100 Ifit1* Ifit3*
Ifitm3* X Gbp2b* 2 50 spot counts in ELISPOT
Ifi203* γ Ifi205* 1 (n = 4)
Tgtp1* IFN- Bst2* 0 Rsad2* cKO SD MT 0 Cxcl9* = 7) = 7) = 13) = 12) (n (n = 10) = 12) Cxcl10* −1 (n (n (n (n Mx1*
Ccl5* Tsc22d3- −2 Ly6c1* WT SD -cKO SD b g WT SD MTXcKO SD MTX Calr Ctsh Ctsb Ctso - * −2 −1 012 Tap2 WT SD anti-PD1 cKO SD anti-PD-1 Lamp1 Tsc22d3 - Sec61 Sec61 * * Tsc22d3 Tsc22d3
1438 NATURE MEDICINE | VOL 25 | SEPTEMBER 2019 | 1428–1441 | www.nature.com/naturemedicine NATURE MEDICINE ARTICLES
a bcSerum *** P = 0.0005 * P = 0.0454 Serum MIFE P = 0.026 ** P = 0.001 *** P < 0.0001 500 500 600 P = 0.041 300 250 250 12 ) 120 150 100 8 80 relative expression 100 in PBMC relative
expression P = 0.39 4 60 Cortisol (ng/ml 4 50 40 P = 0.309 TSC22D3 20 TSC22D3 0 0 0 1 2 1 2 Colon Colon Lung Lung Healthy Healthy cancer cancer Healthy Colorectal cancer Healthy cancer cancer Colorectalcancer volunteer volunteer volunteers volunteers d e f Colorectal cancer Lung cancer cohort 1 Colorectal cancer Lung cancer Lung cancer cohort 2 Lung cancer Gastric cancer 300 200 3 2 Spearman r = 0.2856 r = 0.0456 6 Spearman r = 0.5515 P = 0.0181 P < 0.0001 150 P = 0.0492 ) ) 200 2 in TIDCs
100 3
100 1 Cortisol (ng/ml Cortisol (ng/ml
50 TSC22D relative expression ( × 10,000) 0 0 0 50 75 100 125 150 0 200 400 600 0 5 10 15 POMS score TSC22D3 in PBMCs TSC22D3 in PBMCs relative expression relative expression (×10,000)
g hi Lung cancer (smokers) Colorectal cancer Gastric cancer 100 100 100 HR = 1.45 (1.14–1.85) HR = 1.512 (1.13–2.02) HR = 1.21 (1.01–1.46) log-rank P = 0.0021 log-rank P = 0.0053 log-rank P = 0.042 80 80 80
60 60 60
Probability 40 Probability 40 Probability 40
TSC22D3 TSC22D3 TSC22D3 20 20 Low (n = 253) 20 Low (n = 241) Low (n = 288) High (n = 567 ) High (n = 241) High (n = 588) 0 0 0 0 50 100 150 200 0 50 100 150 200 0 50 100 150 Time (months) Time (months) Time (months)
Fig. 6 | Prognostic value of TSC22D3 in patients with cancer. a, Plasma cortisol levels in healthy volunteers (n"="258) and patients with colorectal (n"="149) or lung (n"="152) cancer were detected by ELISA. An unpaired two-tailed t test with Welch’s correction was used for statistical analyses (P values are shown between the indicated groups). b, TSC22D3 expression in PBMCs from healthy volunteers (n"="61) and patients with colorectal (n"="96) or lung (n"="61) cancer was quantified with qRT–PCR, using GAPDH as an internal control. An unpaired two-tailed t test with Welch’s correction was used for statistical analyses (P values are shown between the indicated groups). c, PBMCs from two healthy volunteers were cultured in plasma from two patients with colon cancer or another two healthy volunteers with or without MIFE. TSC22D3 expression was measured by qRT–PCR with duplicates, using GAPDH as an internal control. A two-tailed Mann–Whitney U test was used for statistical analyses (P values are shown between indicated groups). d, Linear regression tests were applied to explore the correlation between plasma cortisol levels and the POMS score in patients with colorectal (n"="69) or lung (n"="51) cancer. e, Spearman’s rank correlation test was applied to evaluate the link between plasma cortisol levels and TSC22D3 expression by PBMCs from patients with colorectal (n"="70), gastric (n = 50) or lung (n"="62) cancer. f, Spearman’s rank correlation test was applied to analyze the link between the relative expression of TSC22D3 in TIDCs (calculated as the ratio of TSC22D3 UMI counts in TIDCs to total UMI counts in TIDCs) and the relative expression of TSC22D3 in PBMCs (calculated as the ratio of TSC22D3 UMI counts in PBMCs to total UMI counts in PBMCs) in individual patients with lung cancer (n"="5 for cohort 1 from Harvard Medical School49 and n"="5 for cohort 2 from the Sun Yat-Sen University Cancer Center). g–i, The OS of patients with lung cancer (smokers in g), with colorectal cancer (h) or with gastric cancer (i) was compared between individuals bearing tumors with high or low levels of TSC22D3 mRNA transcription, as measured by microarray or RNAseq. Hazard ratio, 95% confidence interval and P values from the log-rank test are shown. Data from qRT–PCR are shown as mean"±"s.d. The other dot plots are shown as means"±"s.e.m. *P < 0.05, **P < 0.01, ***P < 0.001.
Received: 15 April 2018; Accepted: 31 July 2019; 2. Font-Burgada, J., Sun, B. & Karin, M. Obesity and cancer: the oil that feeds Published online: 9 September 2019 the fame. Cell Metab. 23, 48–62 (2016). 3. Dagogo-Jack, I. & Shaw, A. T. Tumour heterogeneity and resistance to cancer References therapies. Nat. Rev. Clin. Oncol. 15, 81–94 (2018). 1. Schoemaker, M. J. et al. Psychological stress, adverse life events and breast 4. Batty, G. D., Russ, T. C., Stamatakis, E. & Kivimaki, M. Psychological distress cancer incidence: a cohort investigation in 106,000 women in the United in relation to site specifc cancer mortality: pooling of unpublished data from Kingdom. Breast Cancer Res. 18, 72 (2016). 16 prospective cohort studies. BMJ 356, j108 (2017).
NATURE MEDICINE | VOL 25 | SEPTEMBER 2019 | 1428–1441 | www.nature.com/naturemedicine 1439 ARTICLES NATURE MEDICINE
5. Hanahan, D. & Weinberg, R. A. Te hallmarks of cancer. Cell 100, 35. Gyorfy, B., Surowiak, P., Budczies, J. & Lanczky, A. Online survival analysis 57–70 (2000). sofware to assess the prognostic value of biomarkers using transcriptomic 6. Galluzzi, L., Buque, A., Kepp, O., Zitvogel, L. & Kroemer, G. Immunological data in non-small-cell lung cancer. PloS One 8, e82241 (2013). efects of conventional chemotherapy and targeted anticancer agents. 36. Szasz, A. M. et al. Cross-validation of survival associated biomarkers in Cancer Cell 28, 690–714 (2015). gastric cancer using transcriptomic data of 1,065 patients. Oncotarget 7, 7. Ma, Y. et al. Anticancer chemotherapy-induced intratumoral recruitment and 49322–49333 (2016). diferentiation of antigen-presenting cells. Immunity 38, 729–741 (2013). 37. Taker, P. H. et al. Chronic stress promotes tumor growth and angiogenesis 8. Ma, Y., Pitt, J. M., Li, Q. & Yang, H. Te renaissance of anti-neoplastic in a mouse model of ovarian carcinoma. Nat. Med. 12, 939–944 (2006). immunity from tumor cell demise. Immunol. Rev. 280, 194–206 (2017). 38. Cui, B. et al. Stress-induced epinephrine enhances lactate dehydrogenase 9. Vacchelli, E. et al. Chemotherapy-induced antitumor immunity requires A and promotes breast cancer stem-like cells. J. Clin. Invest. 129, formyl peptide receptor 1. Science 350, 972–978 (2015). 1030–1046 (2019). 10. Gonzalo, J. A., Gonzalez-Garcia, A., Martinez, C. & Kroemer, G. 39. Bereshchenko, O. et al. GILZ promotes production of peripherally induced Glucocorticoid-mediated control of the activation and clonal deletion of Treg cells and mediates the crosstalk between glucocorticoids and TGF-beta peripheral T cells in vivo. J. Exp. Med. 177, 1239–1246 (1993). signaling. Cell Rep. 7, 464–475 (2014). 11. Michaud, K., Matheson, K., Kelly, O. & Anisman, H. Impact of stressors in a 40. Hamdi, H. et al. Induction of antigen-specifc regulatory T lymphocytes by natural context on release of cortisol in healthy adult humans: a meta- human dendritic cells expressing the glucocorticoid-induced leucine zipper. analysis. Stress 11, 177–197 (2008). Blood 110, 211–219 (2007). 12. Cain, D. W. & Cidlowski, J. A. Immune regulation by glucocorticoids. 41. Futterleib, J. S., Feng, H., Tigelaar, R. E., Choi, J. & Edelson, R. L. Activation Nat. Rev. Immunol. 17, 233–247 (2017). of GILZ gene by photoactivated 8-methoxypsoralen: potential role of 13. Webster, J. I., Tonelli, L. & Sternberg, E. M. Neuroendocrine regulation of immunoregulatory dendritic cells in extracorporeal photochemotherapy. immunity. Ann. Rev. Immunol. 20, 125–163 (2002). Transfus. Apher. Sci. 50, 379–387 (2014). 14. Chavan, S. S., Pavlov, V. A. & Tracey, K. J. Mechanisms and therapeutic 42. Li, C. C., Munitic, I., Mittelstadt, P. R., Castro, E. & Ashwell, J. D. Suppression relevance of neuro-immune communication. Immunity 46, 927–942 (2017). of dendritic cell-derived IL-12 by endogenous glucocorticoids is protective in 15. Golden, S. A., Covington, H. E. 3rd, Berton, O. & Russo, S. J. A LPS-induced sepsis. PLoS Biol. 13, e1002269 (2015). standardized protocol for repeated social defeat stress in mice. Nat. Protoc. 6, 43. Norbiato, G., Bevilacqua, M., Vago, T. & Clerici, M. Glucocorticoids 1183–1191 (2011). and interferon-alpha in the acquired immunodefciency syndrome. 16. Papadopoulou, A. et al. Acute and chronic stress diferentially regulate J. Clin. Endocrinol. Metab. 81, 2601–2606 (1996). cyclin-dependent kinase 5 in mouse brain: implications to glucocorticoid 44. Tormann, V. et al. Genomic dissection of enhancers uncovers principles of actions and major depression. Transl. Psychiatry 5, e578 (2015). combinatorial regulation and cell type-specifc wiring of enhancer-promoter 17. Obeid, M. et al. Calreticulin exposure dictates the immunogenicity of cancer contacts. Nucleic Acids Res. 46, 2868–2882 (2018). cell death. Nat. Med. 13, 54–61 (2007). 45. Navari, R. M. & Aapro, M. Antiemetic prophylaxis for chemotherapy-induced 18. Spitz, I. M. & Bardin, C. W. Mifepristone (RU 486)-a modulator of progestin nausea and vomiting. N. Engl. J. Med. 374, 1356–1367 (2016). and glucocorticoid action. N. Engl. J. Med. 329, 404–412 (1993). 46. Cheng, K. K. F., Lim, Y. T. E., Koh, Z. M. & Tam, W. W. S. Home-based 19. Fleseriu, M. et al. Mifepristone, a glucocorticoid receptor antagonist, multidimensional survivorship programmes for breast cancer survivors. produces clinical and metabolic benefts in patients with Cushing’s syndrome. Cochrane Database Syst. Rev. 8, CD011152 (2017). J. Clin. Endocrinol. Metab. 97, 2039–2049 (2012). 47. Bradt, J., Dileo, C., Magill, L. & Teague, A. Music interventions for improving 20. Bai, Y. Y. et al. ProBDNF signaling regulates depression-like behaviors psychological and physical outcomes in cancer patients. Cochrane Database in rodents under chronic stress. Neuropsychopharmacology 41, Syst. Rev. https://doi.org/10.1002/14651858.CD006911.pub3 (2016). 2882–2892 (2016). 48. Tritos, N. A. & Biller, B. M. Advances in medical therapies for Cushing’s 21. David, D. J. et al. Neurogenesis-dependent and -independent efects syndrome. Discov. Med. 13, 171–179 (2012). of fuoxetine in an animal model of anxiety/depression. Neuron 62, 49. Zilionis, R. et al. Single-cell transcriptomics of human and mouse lung 479–493 (2009). cancers reveals conserved myeloid populations across individuals and species. 22. Cryan, J. F., Page, M. E. & Lucki, I. Diferential behavioral efects of the Immunity 50, 1317–1334 e1310 (2019). antidepressants reboxetine, fuoxetine, and moclobemide in a modifed forced swim test following chronic treatment. Psychopharmacology 182, Acknowledgements 335–344 (2005). Y.M. is supported by China Ministry of Science and Technology (National key research 23. Roni, M. A. & Rahman, S. Efects of lobeline and reboxetine, fuoxetine, or and development program, Grant 2017YFA0506200), Natural Science Foundation of bupropion combination on depression-like behaviors in mice. Pharmacol. China (NSFC) Grant 81972701, 81722037 and 81671630), Natural Science Foundation Biochem. Behav. 139, 1–6 (2015). of Jiangsu Province (Grant BK20170006 and BK20160379), CAMS Innovation Fund for 24. Oh, K. S. et al. Anti-infammatory chromatinscape suggests alternative Medical Sciences (CIFMS; 2016-I2M-1-005 and 2019-I2M-1-003), Non-profit Central mechanisms of glucocorticoid receptor action. Immunity 47, Research Institute Fund of Chinese Academy of Medical Sciences (2017NL31004 298–309 e295 (2017). and 2017RC31008), National Thousand Talents Recruitment Program (China) and 25. Ayroldi, E. et al. Modulation of T-cell activation by the glucocorticoid- Innovative and Entrepreneurial Team Program (Jiangsu Province). G.K. and L.Z. induced leucine zipper factor via inhibition of nuclear factor kappaB. Blood are supported by the Recruitment Program of High-end Foreign Experts in China 98, 743–753 (2001). (GDW20171100085, GDW20181100051). Y.H. is supported by NSFC Grant 81802870 26. Ronchetti, S., Migliorati, G. & Riccardi, C. GILZ as a mediator of and Natural Science Foundation of Jiangsu Province Grant BK20170407. Z.S.Q. is the anti-infammatory efects of glucocorticoids. Front. Endocrinol. 6, supported by NSFC Grant 31600959 and Natural Science Foundation of Jiangsu Province 170 (2015). Grant BK20160378. G.K. is supported by the Ligue contre le Cancer (équipe labellisée); 27. Calmette, J. et al. Glucocorticoid-induced leucine zipper enhanced expression Agence National de la Recherche (ANR); Association pour la recherche sur le cancer in dendritic cells is sufcient to drive regulatory T cells expansion in vivo. (ARC); Cancéropôle Ile-de-France; Chancelerie des universités de Paris (Legs Poix), J. Immunol. 193, 5863–5872 (2014). a donation by Elior; European Research Area Network on Cardiovascular Diseases 28. Calmette, J. et al. Glucocorticoid-induced leucine zipper protein controls (ERA-CVD, MINOTAUR); European Union Horizon 2020 (OncoBiome), Fondation macropinocytosis in dendritic cells. J. Immunol. 197, 4247–4256 (2016). Carrefour; Fondation pour la Recherche Médicale (FRM); Institut Gustave Roussy 29. Ngo, D. et al. Divergent efects of endogenous and exogenous glucocorticoid- (Odysssea), Institut National du Cancer (INCa); INSERM (HTE); Institut Universitaire induced leucine zipper in animal models of infammation and arthritis. de France; the LabEx Immuno-Oncology (ANR-18-IDEX-0001); the RHU Torino Arthritis Rheumatol. 65, 1203–1212 (2013). Lumière; the Seerave Foundation; the SIRIC Stratified Oncology Cell DNA Repair 30. Deng, L. et al. STING-dependent cytosolic DNA sensing promotes and Tumor Immune Elimination (SOCRATE); and the SIRIC Cancer Research and radiation-induced type I interferon-dependent antitumor immunity in Personalized Medicine (CARPEM). J. Calmette was supported by a doctoral fellowship immunogenic tumors. Immunity 41, 843–852 (2014). from the ‘Innovation Thérapeutique du fundamental à l’appliqué’ Doctoral School (ED 31. Sistigu, A. et al. Cancer cell-autonomous contribution of type I interferon 569), LabEX LERMIT. We acknowledge C. Sévoz-Couche and C. Brouillard (UMR_S signaling to the efcacy of chemotherapy. Nat. Med. 20, 1301–1309 (2014). 1158, Unité de Neurophysiologie Respiratoire Expérimentale et Clinique, Sorbonne 32. Grove, R. & Prapavessis, H. Preliminary evidence for the reliability and Université), and P. Li and J. Zhu (Suzhou Institute of Systems Medicine, Chinese validity of an abbreviated profle of mood states. Int. J. Spot Psychol. 23, Academy of Medical Sciences & Peking Union Medical College) for technical assistance. 93–109 (1992). We thank F. Bachelerie for access to an animal facility (UMS/IPSIT, Clamart). 33. Baker, F., Denniston, M., Zabora, J., Polland, A. & Dudley, W. N. A POMS short form for cancer patients: psychometric and structural evaluation. Author contributions Psychooncology 11, 273–281 (2002). Y.M. and G.K. designed this study. Y.M., H.Y., L.X. and S.L. performed experiments 34. Li, T. et al. TIMER: A web server for comprehensive analysis of tumor- involving murine tumor models. Jian Chen and M.S. performed the POMS assessment infltrating immune cells. Cancer Res. 77, e108–e110 (2017). on patients with cancer. Jian Chen, M.S., Z.P., L.F., J.Y. and P.Z. collected blood samples
1440 NATURE MEDICINE | VOL 25 | SEPTEMBER 2019 | 1428–1441 | www.nature.com/naturemedicine NATURE MEDICINE ARTICLES from patients and healthy volunteers. L.X., S.L., V.M. and Y.M. performed behavioral Additional information tests. L.L. provided quality control for some behavioral tests. L.X., Y.M., H.Y., S.L. Extended data is available for this paper at https://doi.org/10.1038/s41591-019-0566-4. and Q.L. performed FACS and RNA-seq analyses. L.X., Y.M. and S.Z. performed Supplementary information is available for this paper at https://doi.org/10.1038/ ELISA and ELISPOT. Q.L. and Y.M. performed survival analyses using a public cancer s41591-019-0566-4. database. B.K., Z.Z., A.L., K.Y. and J.H. analyzed single-cell-sequencing datasets. L.X., S.Z., Q.L., S.L., M.L. and Jinfeng Chen performed cell culture, qRT–PCR and Reprints and permissions information is available at www.nature.com/reprints. immunohistochemistry experiments. J. Calmette, E.M., G.S.-L. and R.K. provided Correspondence and requests for materials should be addressed to G.K. or Y.M. support in the experiments using Tsc22d3-Tg and Tsc22d3-cKO mice. G.K. and Y.M. Peer review information: Kate Gao was the primary editor on this article and managed wrote the manuscript with help of L.Z. its editorial process and peer review in collaboration with the rest of the editorial team. Competing interests Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in L.Z. and G.K. are founders and shareholders of EVERIMMUNE. The other authors published maps and institutional affiliations. declare no competing interests. © The Author(s), under exclusive licence to Springer Nature America, Inc. 2019
NATURE MEDICINE | VOL 25 | SEPTEMBER 2019 | 1428–1441 | www.nature.com/naturemedicine 1441 ARTICLES NATURE MEDICINE
Methods their home cages. Blood samples was collected at the indicated time points (before Mice. Male C57BL/6 and FVB mice aged 7–8 weeks (weighing 21–23 g) and male stress, immediately after stress, and 2 h after stress) and stored at − 80 °C before CD-1 retired breeder mice (aged 6–7 months) were purchased from Charles River further analysis. (Beijing, China). All mice were maintained in specifc-pathogen-free conditions under 12-h/12-h light/dark cycles, with access to food and water ad libitum. Urethane-induced lung cancer model. Naive male FVB mice received Tsc22d3-Tg (CD11c-GILZhi) mice bearing a CD11c (Itgax)-promoter-driven single intraperitoneal (i.p.) injections of urethane (dissolved in 0.9% NaCl saline solution, targeted insertion of Tsc22d3 (Gilz) coding sequences on the X chromosome were 1 g per kg of body weight) once per week, for a total of 6 weeks. Mice were generated and backcrossed to the C57BL/6 background27. Tis strain was created exposed to repeated daily SD stress between days 7 and 16. Blood samples were and maintained in the animal facility of INSERM UMR_S996, Clamart, France. collected before and after SD procedures. On days 23, 25, 50 and 52 after the first Hemizygous male mice and wild-type littermate controls aged 7–10 weeks were injection of urethane, mice were treated with either OXA (5 mg per kg of body used in the experiments. Tsc22d3-cKO (CD11c-GILZKO) mice were generated by weight, from MedChem Express) or the solvent control. On days 111 and 112, crossing Itgax-Cre with Gilzfox/fox mice28,29,50, homozygous mice aged 7–10 weeks the presence of neoplastic nodules were examined with whole-body U-SPECT+/ were used in this study. Tese strains were obtained under MTA authorizations CT scan, a preclinical nuclear imaging system developed by MILabs (MILabs, The from INSERM and Monash University, and were maintained in the animal facility Netherlands) when mice were anesthetized with inhalational isoflurane. The voxel of Suzhou Institute of Systems Medicine. Any requests for these mouse strains resolution applied was 2.5 μm. All mice were euthanized between days 120 and should obtain MTA approvals from INSERM and/or MONASH University. 124, and lung lobes were taken for the neoplasia quantification and histochemical All experimental protocols were approved by the Ethics Committee of Chinese analyses. The number of neoplasias was quantified with a trinocular plain Academy of Medical Sciences and International Animal Care and Use Committee stereomicroscope (SMZ168TP, Motic, Hongkong) equipped with a high-resolution of Suzhou Institute of Systems Medicine (no. ISM-IACUC-0001). camera (Sony HD 1/1.8 inch color CCD). Image capture and analyses were performed with EZ-NET software. Repeated social defeat model. Male CD-1 aggressor mice were allowed to habituate individually in their home cage for 7 d prior to experiments. C57BL/6J Colitis-associated colorectal cancer model. Naive male C57BL/6 mice received ‘screener’ male mice were used to test the aggressiveness of CD-1 mice. During 2 two i.p. injections of 8 mg azoxymethane (AOM, from Sigma Aldrich) per kg of consecutive 3-min screening sessions, the latency to initial physical attacks must body weight on days 0 and 7. To imitate inflammation-driven tumor progression, have been <60 sec15. Moreover, the CD-1mouse must have continuously attacked these mice then received 3 rounds of 1-week 1.5% dextran sodium sulfate (DSS, the C57BL/6J screener mouse for at least 10 sec without interruption for all the 3 of from MP Biomedicals) treatment in drinking water, with a 2-week interval after the 5-min sessions. To start the social defeat protocol, ‘qualified’ CD-1 mice were each DSS treatment. SD procedures were performed daily between days 0 and 9. preconditioned daily in a dark room with red light for 30 min before encountering On days 52 and 55 after the first injection of AOM, mice were treated with 5 mg intruder mice. A C57BL/6 mouse was introduced into the home cage of a CD-1 OXA per kg of body weight. On day 51 (1 d before the antitumor treatment) and mouse for 10 min. Attacked intruder mice usually displayed subordinate posturing. days 87–90, the presence of neoplastic nodules in mice was assessed by whole-body Physical injury was carefully avoided to circumvent biased observations caused U-SPECT+/CT scan (MILabs, The Netherlands) with a voxel resolution of 2.5 μm. by wound healing, inflammation or infection. The 2 mice were then physically Mice were anesthetized with inhalational isoflurane during the micro-computed separated with a perforated plexiglas divider in the center of cage, allowing for a tomography (micro-CT) scan. All mice were euthanized on day 55 or day 90 so we 24-h sensory interaction, with ad libitum access to water and food. To prevent any could quantify the number and size of neoplastic nodules in colon and rectum with habituation to the CD-1 mice, each intruder C57BL/6 mouse was exposed to a a stereo microscope (Motic SMZ168TP) and the supporting softwa.re EZ-NET. new aggressor daily for 10 consecutive days (10 min of physical contact, followed separation on either side of the plexiglas)51,52. Control mice were housed in pairs Chemotherapy, tumor vaccine and immunotherapy. For the establishment of (both C57Bl/6) in cages with a plexiglas divider. The amounts of food and water syngeneic solid tumors, CB57BL/6 mice received subcutaneous (s.c.) injection of 6 consumed, as well as body weight, were measured each morning throughout the 1 × 10 MCA205, TC-1 or CT26. Tumor growth was monitored routinely with an 2 entire SD procedures for each individual mouse. Behavioral tests, as described electronic caliper. When tumor size reached 25–45 mm , mice were treated with below, and blood-sample collection were performed at the indicated time points. either 5.2 mg MTX per kg body weight (i.p., in 200 μl PBS, 1 injection), 5 mg OXA per kg body weight (i.p., in 200 μl PBS, 1 injection), 2.9 mg DX per kg body weight Behavioral tests. Light–dark box tests were carried out in a plexiglas apparatus (intratumoral, in 50 μl PBS, 1 injection) or 10 mg anti-PD-1 mAb (clone RMP1-14, containing a dark chamber (18 cm × 27 cm × 30 cm) and an illuminated (∼200 lux) BioXCell) per kg body weight (intravascularly, in 100 μl PBS, 3 injections every chamber (27 cm × 27 cm × 30 cm). These two compartments were connected with a other day). Control mice received equivalent volumes of solvent or isotype Ab door (6.5 cm × 6.5 cm). Mice were placed in the corner of the dark chamber facing (clone 2A3, BioXCell) through matched injection routes. For tumor vaccination, away from the door and then released. Mice were allowed to explore the arena tumor cells were pretreated with 2 μM mitoxantrone for 16 h, washed with PBS 6 freely for 10 min. A video-tracking system and SuperMaze V2.0 software (Xinruan and injected s.c. (1 × 10 tumor cells in 100 μl PBS) into naive CB57BL/6 mice. 6 Information Technology Co. Ltd, Shanghai, China) were used to record and Ten days later, these mice were rechallenged with 1 × 10 live tumor cells. Tumor analyze the position and locomotion of each mouse, including total locomotion, growth was recorded every 2–3 d. Tumor-bearing mice were euthanized if tumor 2 locomotion in the light box and times of entry and duration of stay in the light box. size reached 300 mm . The open-field test was performed in 4 open boxes (50 cm × 50 cm each) For chemotherapy and immunotherapy experiments, SD procedures were simultaneously, which were separated by plexiglas walls. Mice were placed in the performed daily between days − 12 and − 2, and were followed by behavioral tests corner of each box and left to explore the arena for 1 h under low-light conditions and tumor-cell implantation. MIFE was purchased from Sigma Aldrich, dissolved (5 lux). The position and locomotion of each mouse were recorded and analyzed in double-distilled water (ddH2O) and applied i.p. at a dose of 4 mg per kg of with either an automatic detection system using infrared floor (ViewPoint, Lyon, body weight on days 3, 4, 5, 6, 7, 8, 10, 12 and 14 after tumor-cell injection, or France), or with a video tracking system with high-resolution digital cameras daily for 7 consecutive days following tumor vaccination. FLU and REX (from (SuperMaze V2.0, Shanghai, China). To detect the exploratory behaviors in MedChemExpress) were dissolved in 0.9% NaCl solution and applied i.p. for 14 the center zone, a virtual zone (20 cm × 20 cm) was predefined in the center of consecutive days (20 mg FLU per kg of body weight and 10 mg REX per kg of body each box. All tests were performed within 48 h after the last defeat session. All weight) since the day of tumor-cell injection. The 3-d acute restraint procedure was boxes were wiped with wet cotton pads and then dry cotton pads to remove any performed starting from 1 d before chemotherapy. DEX (3 daily injections, 100 μg, remaining smell, urine and feces before each new test. 20 μg or 5 μg per 20 g body weight) or PRED (3 daily injections, 100 μg or 20 μg per Social-avoidance tests were performed in the same open-field boxes to which 20 g body weight) were administrated intravascularly in non-stressed mice starting the mice previously habituated under low-light conditions (5 lux). In the first from 1 d before MTX. 150-sec session, mice were allowed to explore the open field with an empty round wire cage (18 cm × 9 cm) located at the middle of one side. In the second 150-sec Flow cytometry. Freshly recovered tumors, spleens or draining lymph nodes session, the empty cage was replaced with a cage containing a CD-1 aggressor were cut into small pieces by surgical scissors, or were dissociated with tweezers mouse (referred to as the ‘target’). The interaction between a test subject and in digestion buffer, serum-free RPMI Medium 1640-GlutaMAX-I containing 0.4 CD-1 aggressor was quantified by the time spent in a virtual interaction zone (area Wünsch units per ml Liberase TL (Roche, Mannheim, Germany) and 200 U per projecting 8 cm around wire cage), which was recorded with video tracking systems ml DNase I (Calbiochem, San Diego, CA, USA). These tissues were incubated (ViewPoint, Lyon, France; SuperMaze V2.0, Shanghai, China). The interaction at 37 °C for 30 min (tumor and spleen) or 15 min (lymph nodes), followed by ratio was calculated as: (time spent in the interaction zone in the presence of target transfer to C Tubes for gentleMACS-based dissociation (Miltenyi Biotec) with / time spent in the interaction zone in the absence of target) × 100. m_impTumor_02 and m_impTumor_03 programs. The cell suspension was Splash tests were performed in new cages in dark conditions after 1-h filtered with a 70-μm cell strainers before it was stained with the antibody cocktail. habituation. After receiving two sprays of 10% (wt/vol) sucrose solution, mice were Surface staining was performed with 2 μg per ml of the following fluorochrome- put back in cages to record their grooming behaviors for 5 min53. conjugated antibodies at 4 °C for 25 min. CD45.2 (104), CD11c (N418), F4/80 (BM8), I-A/I-E (M5/114.15.2), CD3 (17A2), CD8a (53-7.6) and IFN-γ (XMG1.2) Acute restraint stress model. Mice were restricted in a 50-ml tube for 1 h per were from BioLegend (San Diego, CA, USA). CD11b (M1/70) and CD45.1 (A20) day (from 9:00 to 10:00) for 3 consecutive days. Control mice were maintained in were from eBioscience, Thermo Fisher Scientific. Ly6G (1A8), Ly6C (AL-21)
NATURE MEDICINE | www.nature.com/naturemedicine NATURE MEDICINE ARTICLES and CD16/32(2.4G2) were from BD Pharmingen (San José, CA, USA). CD16/32 The oval survival, hazard ratio with 95% confidence intervals and log-rank P values (clone 2.4G2) antibody was used to block non-specific staining and the LIVE/ were calculated with a Kaplan–Meier plotter, SurvExpress or GraphPad Prism. DEAD Fixable Yellow Dead Cell Stain Kit from Thermo Fisher Scientific was used to exclude dead cells from further analyses. Cell Proliferation Dye eFluor Gene expression analysis. Total RNA from DCs, macrophages, neutrophils 670 was from eBioscience, Thermo Fisher Scientific. All samples were examined and T cells from tumors, spleens or lymph nodes, as well as from BMDCs, was with Attune NxT Flow Cytometer (Thermo Fisher Scientific) or BD LSR II Flow extracted using RNeasy Micro Kits (Qiagen, Hilden, Germany), following the Cytometer and analyzed with FlowJo software (Tree Star, Inc., Ashland, OR, USA). manufacturer’s instructions. The concentration and quality of RNA samples were Alternatively, DC, macrophage, neutrophils and T cells were isolated with the determined by the NanoDrop 2000 spectrophotometer (NanoDrop technologies, FACSAria III cell sorter (Becton Dickinson, San José, CA, USA) following a pre- USA) and a RNA LabChip 6000 Nano kit (Agilent Technologies, USA). DC enrichment procedure with mouse CD45 microbeads (Miltenyi Biotec). samples with a total RNA concentration above 50 ng were used for RNA-seq (GENEWIZ, Suzhou, China). Raw sequencing data are available at the NCBI Frozen tissue section and immunofluorescent staining. Freshly collected Sequence Read Archive (accession number: PRJNA555789, PRJNA556553). Data tumors or lung lobes were treated in 4% PFA for 4 h at room temperature (RT), analysis was performed with DESeq2, pheatmap and clusterProfiler packages using or overnight at 4 °C. PFA-fixed tissues were transferred to 30% (wt/vol) sucrose R and R studio software (R Foundation for Statistical Computing). Significantly solution (in 1× PBS) for 24 h, embedded in optimum cutting temperature upregulated and downregulated genes (fold change ≥ 1.5 and P < 0.05) were compound and then frozen at − 80 °C. Frozen tissue sections (5 μM) were obtained used to generate a gene ontology network, which was visualized using the with the CM1950 Cryostats (Leica Biosystems), adhered to poly-l-lysine-coated BiNGO (v3.0.3) and EnrichmentMap (v2.1.0) plugins in Cytoscape software55 slides and stored at − 80 °C. Frozen tissue sections were recovered to RT and (v3.5.1). Nodes represent enriched GO terms, whereas node size represents the washed three times before histochemistry analysis with an H&E staining kit corresponding false discovery rate (FDR)-adjusted enrichment P value (q value). (Sangon Biotech, China). Alternatively, samples were fixed with 2% PFA for 15 min, Ingenuity pathway analysis software (Qiagen) was used to explore the expression permeabilized with 0.5% Triton X-100 for 5 min, blocked with 20% FBS for 20 min of genes downstream of Tsc22d3. Their expressions in tumor-infiltrating DCs were and stained with 5 μg per ml fluorescent-dye-conjugated mouse antibody (CD3, summarized from RNA-seq data and plotted in a heat map. clone 17A2; IFN-γ, clone XMG1.2) for 2 h at RT. Tissue sections were then washed Bone marrow cells were cultured in complete DMEM containing 10% with 1× PBS, followed by nuclear counterstaining with chromosomal dye Hoechst FBS, mouse GM-CSF and IL-4 (20 ng ml–1 for each cytokine, Miltenyi Biotec). 33258 (2 μg per ml, 5 min at RT). After 3 thorough washes with 1× PBS, slides were The purity of BMDCs was verified with CD11c (clone N418), I-A/I-E (clone sealed with Fluoromount-G mounting medium (Southern Biotech). Images were M5/114.15.2) and CD86 (clone GL-1) antibodies (from BD bioscience or captured with confocal microscope (Leica TCS SP8, installed at Suzhou Institute Biolegend) on day 7 after initial culture. Splenic T cells were purified with mouse of Systems Medicine, China). For each group, at least 2 non-consecutive tissue Pan T Cell Isolation Kit II (Miltenyi Biotec). BMDCs or splenic T cells were sections were stained for each sample, and around 100 view fields were recorded cultured with live or MTX-treated dying MCA205 cells in the presence of DEX for quantification. (10− 7 M), or with serum freshly collected from Ctrl or SD- or AR-preconditioned mice in the presence of the glucocorticoid receptor antagonist MIFE overnight. Clinical investigations and data analysis. Clinical investigations in patients The amount of MIFE was adjusted to ten times the corticosterone level of each with cancer and healthy volunteers were approved by the Ethics Committee of serum sample. The expression of Tsc22d3 was detected by qRT–PCR. Nantong Tumor hospital (K-2017-001) and Sun Yat-Sen University Cancer Center For quantitative qRT–PCR, total RNA (100 ng) from mouse BMDCs, human (GZR2017-216). We have obtained informed consent from all participants. Blood PBMCs or mouse tumor-infiltrating DCs, T cells, macrophage and neutrophils samples were collected in early-morning fasting states from patients with cancer were reversely transcribed with Oligo(dT)20 primers using Superscript III reverse and healthy volunteers, with BD Vacutainer plastic EDTA tubes. Plasma and transcriptase (Thermo Fisher Scientific, USA), according to the manufacturer’s PBMCs were freshly isolated with Human Lymphocyte Separation Tube (Dakewe instructions. Gene expression assays were performed with Universal Master Mix II Biotech Co., Shenzhen, China). Plasma samples were aliquoted and frozen at (with UNG) (Applied Biosystems, Foster City, CA, USA) and appropriate Taqman − 80 °C. PBMCs were resuspended in RZ buffer from RNAsimple Total RNA kit probes (Thermo Fisher Scientific). The murine housekeeping gene peptidylprolyl (TIANGEN BIOTECH, Beijing, China) and stored immediately at − 80 °C. Patients isomerase A (Ppia) was used as control for data normalization. Fold changes with cancer who received glucocorticoids concomitantly with chemotherapy, were calculated using the ΔΔCt method. The expression of human TSC22D3 was or those who might have been exposed to glucocorticoid before blood-sample tested with validated primers (Forward: 5′- GGATGAACAAGGGGATGGCT-3′; collection (owing to CT examination, allergy or other indications) were not Reverse: 5′-ACCCGCTACAGACAAGCTTT-3′) using SYBR Premix EX Taq II included in this study. (TaKaRa Bio Inc., Shiga, Japan). GAPDH expression was used as internal control POMS tests were performed on some hospitalized patients with cancer for data analysis (Forward: 5′-ACAGCCTCAAGATCATCAGC-3′; Reverse: within the 7 d following the blood-sample collection. The 40-item questionnaire 5′-TTCAGCTCAGGGATGACCTT-3′). was designed initially by Grove and Prapavessis32. The tests were performed in a room with ambient temperature, noise level and humidity. And the Visualization and quantification of TSC22D3 expression in scRNA-seq presence of other people was avoided. Patients who participated in these tests datasets. For generating Uniform Manifold Approximation and Projection were not using antidepressant medication. They had not consumed coffee (UMAP) visualization of tumor samples from six published small conditional or tea within the previous 48 h, and their last meals were at least 2 h before RNA (scRNA)-seq studies49,56–60 (Extended Data Fig. 7e), the single-cell expression the test. The questionnaire examined five negative mood states (nervous, angry, matrices were downloaded from the NCBI GEO datasets (GSE103322, GSE114725, fatigue, depressive and panic) and two positive mood states (energetic and GSE120575, GSE127465) and the EBI ArrayExpress database (E-MTAB-6149), self-confident). Each state included a group of questions with 5 rankings and supplemental data were downloaded online. For datasets in which cells were (nothing, 0; mild, 1; moderate, 2; more, 3; and severe, 4). Then the POMS unfiltered, a set of quality-control thresholds (minimum number of expressed value for each patient was calculated as: 100 + (nervous × 0.72) + (angry × genes > 200; proportion of mitochondrial reads < 5%) were applied. Matrices 0.82) + (fatigue × 0.71) + (depressive × 0.66) + (panic × 0.63) – (energetic × 0.62) were then log-transformed and scaled using R package Seurat (https://github. – (self-confident × 0.63). com/satijalab/seurat). For unsupervised clustering analysis, the FindClusters Tumor Immune Estimation Resource (TIMER, https://cistrome.shinyapps. function was applied to the first 20 principle components of each dataset using io/timer/), a web server, was used to explore the expression of TSC22D3 with parameter resolution = 1. Clusters of major immune cell types were annotated the abundance of 6 tumor-infiltrating immune cell types (DCs, neutrophils, according to the expression level of canonical marker genes (T cell: CD3D, CD3E; macrophages, CD4 T cells, CD8 T cells, and B cells) in 23 cancer types in The CD4 T cell: CD4, IL-7R; CD8 T cell: CD8A, CD8B; NK cell: NKG7, NCAM1; Cancer Genome Atlas (TCGA). B cell: MS4A1, CD19; plasma cell: MZB1; CD79A; DC: CD1C, LAMP3, IL-10, CLEC9A; pDC: LILRA4; macrophage: CD68, CD86; mast cell: MS4A2). Two- Overall survival analysis. To verify that the gene expression level of TSC22D3 is dimensional representations were generated with the Seurat function RunUMAP of prognostic value, the datasets of a total of 820 patients with lung cancer who using parameters min.dist = 0.7, spread = 1.2. To generate scatter plots and violin had smoking experience and 876 patients with gastric cancer were extracted from plots, log-normalized expression of TSC22D3 was centered to the mean of zero and Gene Expression Omnibus (GEO) and TCGA databases. These patients were sub- scaled to the variance of one z score for each dataset. grouped into TSC22D3hi and TSC22D3lo, with the best performing threshold to We combined two cohorts of scRNA-seq datasets to explore the correlation separate the low and high quartiles as the cut-off. OS was analyzed with a Kaplan– between TSC22D3 expression in TIDCs and TSC22D3 expression in the matched Meier plotter35. The hazard ratio with 95% confidence intervals and log-rank PBMCs from patients with lung cancer. The first cohort is from a published P values were calculated. For patients with colorectal cancer, meta-analysis of OS was dataset49. The second cohort is from our ongoing study with five participants. performed with SurvExpress54 using quantile normalization of raw data from GEO Only patient samples with >10 TIDCs were included in the analysis. TSC22D3 in database (the GSE12945, GSE14333, GSE17536, GSE17537, GSE31595, GSE41258 TIDC relative expression was calculated as the ratio of unique molecular identifier cohorts). High and low levels of TSC22D3 mRNA transcription were divided at (UMI) counts of TSC22D3 in a TIDC to the total UMI counts (of all genes) in the median of gene expression with the Affymetrix microarray probe ‘207001_x_at’. TIDC population. TSC22D3 in PBMC relative expression was calculated as the Similar analyses were applied for CXCL9 (Affymetrix microarray probe: 203915_ ratio of total TSC22D3 UMI counts to total UMI counts (of all genes) in the bulk of at) and CCL5 (Affymetrix microarray probe: 1405_i_at), using the same datasets. PBMCs. Analyses were performed in R software 3.5.3.
NATURE MEDICINE | www.nature.com/naturemedicine