T-cell–Secreted TNFα Induces Emergency Myelopoiesis and Myeloid-Derived Suppressor Cell Differentiation in Cancer

Steady-state hematopoiesis in the bone marrow is a tightly controlled and regulated process that ensures the continuous generation of all blood lineages (1). In cancer, hematopoiesis is perturbed and characterized by a preferential myeloid differentiation at the expense of erythroid and lymphoid differentiation (2). This leads to the accumulation of immature and immunosuppressive myeloid cells, primarily myeloid-derived suppressor cells (MDSC; refs. 3, 4). In mice, MDSCs express granulocytic (CD11b⁺Ly6G⁺Ly6C^lo; Gr-MDSC) or monocytic markers (CD11b⁺Ly6C⁺Ly6G^lo; M-MDSCs; ref. 5). They suppress the adaptive immune response to cancer and promote tumor growth by promoting tumor cell survival, angiogenesis, and metastasis (4, 5). MDSCs are short-lived and have to be continuously replenished from hematopoietic stem and progenitor cells (HSPC) in the bone marrow and with subsequent mobilization and acquisition of immunosuppressive activity in the tumor microenvironment (5). Although the mechanisms are not yet fully understood, the accumulation of MDSCs and the aberrant myelopoiesis in patients with cancer are attributed to the secretion of tumor-derived factors. Hematopoietic cytokines such as GM-CSF, G-CSF, IL6, and IL1 are produced in a variety of human tumors such as brain, colorectal, and lung cancer and regulate the production of MDSCs from bone marrow progenitors (6–8). In the MMTV-PγMT breast cancer mouse model, G-CSF released by mammary tumor cells induced hematopoietic stem cell (HSC) expansion and granulopoiesis in the bone marrow to replenish short-living MDSCs (7, 9). Similarly, it has been documented that tumor growth in Lewis lung carcinoma model is accompanied with an increase in peripheral myeloid cells and lineage (Lin)⁻c-kit⁺sca-1⁺ stem and progenitor cells (LSK). This was attributed to insulin-like growth factor-I receptor signaling in HSCs (10). Furthermore, GM-CSF has been shown to induce the differentiation of granulocyte monocyte myeloid progenitors (GMP) at the expense of lymphoid and erythroid progenitors (11). Similarly, GM-CSF–secreted by mammary 4T1 tumors led to the expansion of myeloid progenitors and accumulation of CD11b⁺GR-1⁺ myeloid cells (12). In addition, TNFα has been shown to lead to the accumulation of MDSCs in murine and human tumors (13, 14).

Importantly, the vast majority of solid tumors do not secrete hematopoietic cytokines (8). The mechanisms underlying the modulation of myelopoiesis in these tumors are poorly understood. In this study, we document an activated hematopoiesis with increased numbers of long-term (LT) and short-term (ST) HSCs and myeloid progenitor cells in transplanted, chemically induced, and spontaneous murine tumor models. This led to an accumulation of immunosuppressive MDSCs in tumor-bearing mice. Interestingly, TNFα secreted by T cells induced proliferation of HSPCs, myeloid differentiation, and the accumulation of MDSCs. Therefore, the activated adaptive immune system in cancer induces immunosuppressive myeloid cells that dampen the tumor-specific immune response.

C57BL/6 (BL/6), Rag-1^−/− (Rag^−/−), IFNγ-R^−/−, TNFR1/2^−/−, and Ly5.1 mice were from the Institute of Laboratory Animal Science (Zurich, Switzerland). IL6^−/− mice were obtained from M. Kopf (Swiss Federal Institute of Technology, Zurich, Switzerland). Ubi-GFP mice were from C. Müller (Institute of Pathology, University of Bern, Bern, Switzerland). K-ras^LSL-G12D/WT; p53^Fl/Fl (KP) mice were kindly provided by Alfred Zippelius (Tumor Immunology, University of Basel, Basel, Switzerland). All animals were on BL/6 background. All animal experiments were performed in 6- to 8-week-old mice, housed in a specific pathogen-free facility. All animal experiments were approved by the Veterinary Office of the Canton Bern and performed according to Swiss laws for animal protection.

MC57, MC38, B16F10, and 3LL tumors were induced as described by Matter and colleagues (15). Briefly, tumor single-cell suspensions were injected subcutaneously into the flanks of Rag^−/− mice. After 14 days, tumors were collected and nonnecrotic tissue was cut into small fragments (1–2 mm³). Tumor fragments were then transplanted subcutaneously in the flanks of recipient mice. Tumor volume was calculated according to the formula V = π × abc/6, where a, b, and c are orthogonal diameters. For methylcholanthrene (MCA)-induced tumors, 250 μg of MCA dissolved in sunflower oil was injected subcutaneously into shaved flanks of BL/6 mice (control: sunflower oil). For tumor induction in KP transgenic mice, an adenoviral vector expressing Cre recombinase was intratracheally injected into 6-week-old KP mice (16).

MC57 fibrosarcoma, B16F10 melanoma, MC38 colon adenocarcinoma, and mouse Lewis lung carcinoma 3LL cell lines were a gift from Prof. Rolf Zinkernagel, Institute of Experimental Immunology, University of Zurich (Zurich, Switzerland) and have been characterized and described before (15, 17). No additional authentication was performed. Cell cultures were regularly tested for Mycoplasma contamination.

Bone marrow lineage depletion was performed by magnetic-activated cell sorting (MACS) negative selection using biotinylated Abs against red blood cell precursors (α-Ter119), B cells (α-CD19), T cells (α-CD3ϵ), and myeloid cells (α-Gr1), MACS streptavidin beads, and LS columns (Miltenyi Biotec). Negative cell fraction was used for analysis or further cell sorting.

Anti-mouse mAbs against the following antigens were used for flow cytometry: CD4 (GK1.5), CD8 (53-6.7), CD3ϵ (145-2C11), CD19 (6D5), CD11b (M170), Ly6C (HK1.4), Ly6G (1A8), Gr1 (Ly6C/G; RB6-8C5), c-Kit (2B8), CD34 (RAM34), CD16/32 (FcγR; 93), IL-7Rα (CD127; A7R34), CD90.1 (Ox-7), CD90.2 (30-H12), CD48 (HM48-1), CD135 (A2F10), CD150 (TC15-12F12.2), CD45 (30-F11); Sca-1 (D7), CD45.1 (A20), CD45.2 (104), and BrdU and isotype (BD Pharmingen). Cells were washed in PBS and resuspended in the corresponding FACS antibodies for 30 minutes at 4°C. Cells were then washed in PBS and analyzed on a LSRII (BD Biosciences). Alternatively, cells of interest were FACS sorted by FACS Aria II (BD Biosciences). Data were analyzed with FlowJo software (Treestar).

Blood was collected into EDTA-coated tubes and white blood cell counts were determined using a Vet ABC animal blood counter (Medical Solution GmbH) and/or by FACS staining.

Tumors were cut into very small pieces by a scalpel, digested for one hour at 37°C in PBS supplemented with 1 mg/mL Collagenase-IA, 100 μg/mL Hyaluronidase-V (Sigma), 40 U/mL DNase-I (Roche), 5 mmol/L CaCl₂, and 5 mmol/L MgCl₂, washed and filtered to get a single-cell suspension. Tumor-infiltrating lymphocytes (TIL) were isolated by positive MACS of CD45⁺ cells using biotinylated anti-CD45, MACS streptavidin beads, and LS columns (Miltenyi Biotec).

FACS-sorted HSPCs, Lin⁻ bone marrow cells, splenocytes, or blood cells were plated into MethoCult M3134 medium (STEMCELL Technologies) supplemented with 15% FCS, 20% BIT [50 mg/mL BSA in IMDM, 1.44 U/mL rh-insulin (Actrapid; Novo Nordisk), and 250 ng/mL human holo transferrin (Prospec)], 100 μmol/L 2-mercaptoethanol, 100 U/mL penicillin, 100 μg/mL streptomycin, 2 mmol/L l-glutamine, and 50 ng/mL rm-SCF, 10 ng/mL rm-IL3, 10 ng/mL rh-IL6, and 50 ng/mL rm-Flt3-ligand (Prospec). Colonies were counted after 7 days on a DMIL inverted microscope (Leica) equipped with an Intensilight C-HGFI unit (Nikon). For some assays, cells were incubated overnight with 10% sera, T-cell conditioned media (TCM) or tumor cell line–conditioned media before applying to the colony one day later. Blocking antibodies (5 μg/mL) for IL6 (clone MP5-20F3; BioLegend), TNFα (clone MP6-XT22, BioLegend), or CCL3 (clone 39624; R&D) were added to the overnight cell culture where indicated. Control colonies were supplemented with the corresponding isotype controls.

C-kit^hi cells, LSKs, CMPs, and GMPs were sorted on a BD FACS Aria (BD) sorter and incubated in 1% PFA/PBS overnight at 4°C. Samples were permeabilized with 0.2% Triton X-100 for 30 minutes at 4°C and labeled with 5 μg/mL DAPI (Roche).

FACS-sorted CD11b⁺Gr1⁺ MDSCs from tumor-bearing or naïve mice were cultured with anti–CD3ϵ-stimulated T cells from BL/6 mice in a ratio of 3:1, for 3 days. [³H]-Thymidine was added to the culture during the last 16 hours of stimulation. [³H]-Thymidine incorporation was measured using a scintillation beta counter.

Animals were treated with BrdU (Sigma; 0.8 mg/mL in drinking water and 1 mg intraperitoneally/day) on 2 consecutive days and BrdU staining was performed as described in the manufacturer's instructions (BrdU Flow Kit; BD Biosciences).

Recipient mice were lethally irradiated (2 × 6.5 Gy within a 4-hour interval) with a Gamma cell 40 (MDS Nordion). Whole bone marrow cells or CD45⁺ cell from tumors (10⁵ cells) were transplanted along with congenic competitor bone marrow cells (2 × 10⁵ cells) at ratios of 1:2 into recipient mice. During 1 to 2 weeks after transplantation, antibiotics were added to the drinking water.

For T-cell depletion, mice were treated intraperitoneally on day −1, day 0, and then every week after tumor transplantation with 100 μg anti-CD4 antibody (clone GK1.5; Bio X Cell) or anti-CD8 (clone YTS 169.4; Bio X Cell) antibody or both, together with the appropriate isotype control from rat serum. T-cell depletion in blood was controlled by FACS prior to tumor transplantation. Depleting efficiency was higher than 98%. TNFα was neutralized in vivo by administration of anti-TNFα (clone XT3.11; Bio X Cell) or isotype control (clone BE0290; Bio X Cell) twice a week starting at the time point of tumor transplantation.

CD4⁺ or CD8⁺ T cells were sorted by FACS from spleens of naïve or tumor-bearing mice 30 days after tumor transplantation. A total of 3.5 × 10⁵ cells per well were incubated in RPMI 10% FCS for 16 hours at 37°C 5% CO₂. Supernatants were then collected after centrifugation.

Forty-eight mouse cytokines, chemokines, and growth factors were analyzed in sera or TCM using the Multiplexing LASER Bead Assay (Eve Technologies): IL1α, IL12 (p70), IL33, RANTES, IL1β, IL13, Eotaxin, M-CSF, IL2, IL15, IP-10, G-CSF, IL3, IL17A, KC, GM-CSF, IL4, IL17F, LIF, IFNγ, IL5, IL17E/IL25, LIX, TNFα, IL6, IL21, MCP-1, TNFβ, IL7, IL22, MIG, TGFβ1, IL9, IL23, MIP-1α, TGFβ2, IL9, IL27, MIP-1β, TGFβ3, IL10, IL28B, MIP-2, VEGF, IL12 (p40), IL31, MIP-3α, CD40L. GM-CSF was measured in conditioned media of different tumor cell lines. Heatmaps were generated using standard Ward method according to the standard normal distribution.

Canonical pathway representing differentially expressed cytokines were identified using the Ariadne Genomics Pathway Studio software, version 9 (Elsevier). The dataset containing protein (cytokine) names and corresponding fold changes were uploaded into the Pathway Studio. The analysis identified the direct interactions between TNFα and other differentially expressed cytokines.

Statistical analysis was performed using GraphPad Prism 5.0 (GraphPad Software). Data are represented as mean ± SEM. The Shapiro–Wilk test was used to determine whether the data meet the assumption of normality. Data were analyzed using one-way ANOVA and Tukey multiple comparison test, Student t test (two-tailed), one-sample t test or two-way ANOVA, and Bonferroni post hoc test (P value shows interaction). *, P < 0.05 was considered significant; **, P < 0.01, ***, P < 0.001; and ****, P < 0.0001.

To study the mechanisms how solid tumors influence hematopoiesis, we transplanted solid fragments of the fibrosarcoma MC57 subcutaneously into BL/6 mice (15). Recipient mice developed clinically detectable tumors after approximately 1 week that grew up to 1 cm³ within 4 weeks after transplantation (Supplementary Fig. S1A). The analysis of white blood cells in the circulation and in the bone marrow revealed an increase in myeloid cells and a decrease in T-cell numbers (Supplementary Fig. S1B–S1D). Larger tumors often develop central necrosis. This may explain the lack of increase of leukocytes and granulocytes at the end of the experiment. Similarly, frequencies of T and B cells in spleen were reduced, whereas the frequency of myeloid CD11b⁺ cells was increased in tumor-bearing mice (Fig. 1A). CD11b⁺Gr1⁺ MDSC numbers were significantly increased in spleen and bone marrow of tumor-bearing mice compared with naïve mice (Fig. 1B). Importantly, the ratio of T cells to MDSCs was significantly reduced in tumor-bearing mice (Fig. 1C). The number of MDSCs in spleen, bone marrow, and the tumor correlated with the tumor size (Fig. 1D). To test MDSCs functionally in vitro, we stimulated T cells from naïve BL/6 mice with anti-CD3ϵ antibody in the presence of FACS-sorted CD11b⁺Gr1⁺ MDSCs from naïve or tumor-bearing mice. Proliferation of activated T cells, assessed by [³H]-thymidine incorporation, was significantly lower in the presence of MDSCs from tumor-bearing mice compared with controls (Fig. 1E). These data indicate that hematopoiesis in tumor-bearing mice is skewed toward a preferential accumulation of immunosuppressive myeloid cells.

The increase in myeloid cells in tumor-bearing mice depends on an accelerated myelopoiesis in the bone marrow. We therefore performed a detailed analysis of HSPCs in the bone marrow of tumor-bearing mice and naïve controls. The number of lineage-negative (Lin⁻) HSPCs was significantly higher in tumor-bearing mice than in controls (Fig. 2A). Similarly, numbers of LSKs, Lin⁻ sca-1⁻ c-kit⁺ CD34⁺ FcγR⁻ common myeloid progenitors (CMP) and Lin⁻ sca-1⁻ c-kit⁺ CD34⁺ FcγR⁺ GMPs were elevated in tumor-bearing mice (Fig. 2B). However, numbers of Lin⁻ c-kit⁺ CD127⁺ CD90.1/2⁻ common lymphoid progenitors (CLP) were comparable (Fig. 2C). In addition, FACS-purified Lin⁻ cells, LSKs and CMPs from the bone marrow of tumor-bearing mice formed more colonies in methylcellulose than the respective cell populations from control mice (Fig. 2D). A phenotypical subdivision of the LSK population revealed higher numbers of long-term HSCs (LT-HSC, CD34⁻CD48⁻CD135⁻CD150⁺), short-term HSCs (ST-HSC, CD34⁺CD48⁻CD135⁻CD150⁺) and the multipotent progenitors (MPP1, CD34⁺CD48⁺CD135⁻CD150⁺ and MPP2, CD34⁺CD48⁺CD135⁻CD150⁻) in tumor-bearing mice (Fig. 2E). In contrast, numbers of MPP3 (CD34⁺CD48⁺CD135⁺CD150⁻) that are known to be skewed toward lymphoid differentiation (18) remained constant (Fig. 2F). This is in agreement with our observation of comparable numbers of CLPs in both tumor-bearing and naïve mice (Fig. 2C).

Inflammatory stimuli activate and mobilize HSPCs into the circulation and to extramedullary tissues (19, 20). Numbers of Lin⁻ cells increased significantly in the spleen and blood of tumor-bearing animals (Fig. 2G). The increased number of HSPCs in spleen and blood was confirmed functionally by colony-forming assays (Fig. 2H). In addition, LSKs were detected in the tumor tissue (Fig. 2I). Importantly, isolated LSKs from tumors were functional and reconstituted hematopoiesis in lethally irradiated recipient mice similar to bone marrow LSKs isolated from naïve mice (Fig. 2J). In summary, hematopoiesis in tumor-bearing mice is activated with increased mobilization and myeloid differentiation.

To determine whether the elevated numbers of HSPCs are due to enhanced proliferation, we performed a cell-cycle analysis of HSPCs using DAPI staining. C-kit^hi HSPCs, LSKs, and CMPs from tumor mice showed a higher frequency of cells in the replicating S-phase and a lower fraction in the G₁-phase of the cell cycle (Fig. 3A; Supplementary Table S1). GMPs showed a similar trend, however to a lesser extent. In addition, a higher BrdU incorporation in vivo in LSKs and CMPs and a trend to a higher incorporation in GMPs confirmed an enhanced proliferation of HSPCs in tumor-bearing mice (Fig. 3B; Supplementary Table S2). In contrast, there were no significant changes in Annexin-V⁺ cells for CMPs, GMPs, and LSKs in tumor-bearing or naïve mice (Fig. 3C; Supplementary Table S2).

To functionally validate the findings of increased numbers of HSPC in tumor-bearing mice, we transplanted bone marrow cells (Ly5.2) into lethally irradiated Ly5.1 recipient mice (Fig. 3D). In line with our previous results, bone marrow cells from tumor-bearing mice reconstituted primary recipient mice more efficiently compared with bone marrow cells from naïve mice. This was demonstrated by higher percentage of donor Ly5.2⁺ total cells and LSKs in Ly5.1 recipients of bone marrow cells from tumor-bearing rather than naïve mice (Fig. 3E and F). These results functionally confirm a higher number of HSPC in tumor-bearing mice.

We next tested whether the observed changes in myelopoiesis are limited to MC57 fibrosarcoma or if other tumors can similarly activate HSPCs. To this end, we analyzed HSCPs in different murine tumor models. Tumor-bearing mice with MC38 colon carcinoma and B16F10 melanoma did not show significant alterations in the numbers of HSPCs in the bone marrow. Moreover, in vitro assays revealed a comparable colony-forming capacity of HSPCs from tumor-bearing or naïve mice (Fig. 4A–D; Supplementary Fig. S2A and S2B; Supplementary Table S3). In contrast, LSKs and GMPs were increased in mice with 3LL Lewis lung carcinoma (Fig. 4E; Supplementary Table S3; Supplementary Fig. S2C) and HSPCs from tumor mice formed more colonies compared with naïve mice (Fig. 4F).

To mimic a more physiologic situation of tumor development, we inoculated MCA into the flank of BL/6 mice. Mice that developed chemically induced tumors had higher numbers of LSKs and slightly higher numbers of CMPs in bone marrow (Fig. 4G; Supplementary Table S3). This was accompanied by a higher colony formation capacity in vitro (Fig. 4H). Finally, we analyzed bone marrow HSPCs in a genetically engineered mouse model of lung adenocarcinoma. In this model, tumor formation is driven by a conditional overexpression of K-ras^G12D in combination with loss of p53 [K-ras^LSL-G12D/WT; p53^Fl/Fl (KP); ref. 16]. KP mice developed autochthonous lung tumors after inhalation of adenoviral vectors expressing Cre recombinase. Tumor-bearing mice had significantly higher numbers of c-kit^hi cells, LSKs, CMPs and GMPs and HSPCs formed more colonies in vitro compared with nontumor-bearing littermate mice (Fig. 4I and J; Supplementary Table S3). Therefore, HSPC numbers and myelopoiesis are increased in several, but not all tumor models.

Tumors can activate hematopoiesis through the secretion of various cytokines such as GM-CSF and other CSFs (7, 21–23). However, tumor-conditioned media did not significantly enhance colony formation of LSKs (Fig. 5A). In addition, GM-CSF concentrations were not detectable in cultures, except for MC38-conditioned medium (Fig. 5B). Interestingly, although MC38 cells produced detectable levels of GM-CSF, this was not sufficient to activate myelopoiesis in vivo (Fig. 4A).

Alternatively, the inflammatory environment induced by the tumor may indirectly influence HSPCs in the bone marrow. Interestingly, HSPC numbers were increased in tumors that are known to be immunogenic (MC57, 3LL, MCA-induced tumors, and KP lung tumors; refs. 24–28). In contrast, HSPCs remained unchanged in low to nonimmunogenic tumors such as MC38 and B16F10 (29, 30). To investigate a potential role of the adaptive immune system in the activation of the HSPC compartment, we transplanted MC57-tumor fragments into Rag^−/− mice that lack mature T, B, and NKT cells (31). In the absence of the adaptive immune system, numbers of HSPCs and the colony formation capacity did not increase in tumor-bearing mice (Fig. 5C and D). In addition, MDSC numbers in tumor-bearing Rag^−/− mice increased significantly less than in tumor-bearing BL/6 mice (Fig. 5E). These experiments indicate that myelopoiesis in tumor-bearing mice was increased by the adaptive immune system.

To confirm the results observed in the Rag^−/− mice and to analyze which cell population of the adaptive immune system is responsible for the activation of HSPCs, we depleted CD4⁺, CD8⁺, or both T-cell populations before MC57 tumor transplantation in BL/6 mice. Depleting CD4⁺ and CD8⁺ T cells in naïve mice did not change LSK numbers in the bone marrow (Fig. 6A). In contrast, depletion of CD4⁺ and CD8⁺ T cells normalized LSKs and CMPs numbers. Single depletion of CD4⁺ T cells similarly normalized HSPC numbers, whereas depletion of CD8⁺ T cells alone did not (Fig. 6A). In addition, we analyzed the effect of T-cell depletion on MDSC numbers in the spleen. Single and double depletion of CD4⁺ and CD8⁺ T cells in tumor-bearing mice resulted in a significant reduction of MDSC numbers (Fig. 6B). Interestingly, Mo-MDSCs were reduced to a higher extend than Gr-MDSCs (Supplementary Fig. S3A). Taken together, this experiment suggests that mainly CD4⁺ T cells are responsible for increasing HSPC activity and myelopoiesis in tumor-bearing mice. The fact that depletion of CD4⁺ and CD8⁺ T cells comparably reduced MDSC numbers in tumor-bearing mice suggests that CD8⁺ T cells contribute to MDSC differentiation and accumulation by other pathways than regulating hematopoiesis in the bone marrow.

We next analyzed whether the increase in LSK numbers was mediated by a soluble factor secreted by T cells. LSKs formed more colonies in the presence of serum from MC57 tumor-bearing mice compared with serum from naïve BL/6 mice. In contrast, serum from MC57 tumor-bearing Rag^−/− mice did not increase colony formation compared with serum from naïve Rag^−/− mice (Fig. 6C). Interestingly, heat-inactivated serum from MC57 tumor-bearing mice lost its capacity to enhance colony formation (Fig. 6D), indicating that the soluble factor is a protein, most probably a cytokine that is secreted by activated T cells. To confirm our hypothesis, we performed colony assays of LSKs in the presence of conditioned media from T-cell cultures (TCM) originating from MC57 tumor-bearing or naïve mice. CD4⁺ TCM from MC57 tumor-bearing mice significantly increased colony formation capacity of naïve LSKs, whereas CD8⁺ TCM resulted only in a nonsignificant increase in colony numbers (Fig. 6E). A similar increase in colony formation was observed when adding CD4⁺ TCM from mice bearing immunogenic 3LL tumors, but not from mice with less immunogenic MC38 tumors (Fig. 6F). Taken together, these results indicate that CD4⁺ T cells from mice with immunogenic tumors secrete a protein that induces expansion of LSKs.

To define which T cell–derived factors are responsible for the observed activation of HSPCs, we performed a customized array of 48 cytokines, chemokines, and growth factors (listed in Materials and Methods sections) in sera and TCM of tumor-bearing or naïve mice. Analysis of sera from tumor-bearing mice revealed a decreased level of 23 cytokines and an increase in 18 cytokines compared with sera from naïve mice; 7 cytokines were not detected at all (Fig. 6G). In CD4⁺ TCM, 3 cytokines were downregulated and 12 were upregulated; 33 cytokines were not detectable in TCM (Fig. 6H and I). IL6, MIP-1α, and TNFα were among the most significantly upregulated cytokines in CD4⁺ TCMs (Fig. 6J). IL6 is known to induce activation of hematopoiesis with a preferential myeloid differentiation during chronic inflammation (32, 33). MIP-1α is known to promote myeloid differentiation through remodeling the bone marrow niche (34). In contrast, one study indicated that MIP1α is a negative regulator of HSCs (35). TNFα has activating and inhibiting effects on HSPCs depending on its concentration and the presence of other growth factors (36).

To functionally validate whether one of the elevated cytokines is responsible for the activation of HSPCs, we analyzed colony formation of LSKs in the presence of blocking antibodies for IL6, TNFα, and MIP1α. Blocking of IL6 and MIP1α did not reduce the elevated colony formation of LSKs in the presence of serum from tumor-bearing mice. However, blocking TNFα reduced colony formation to the level of control cultures with naïve serum (Fig. 6K). In contrast, addition of TNFα increased the colony formation capacity of LKSs (Fig. 6L). Both, CD4⁺ and CD8⁺ T cells in spleen of tumor-bearing 3LL and MC57 tumors produced TNFα (Supplementary Fig. S3B-C). Importantly, TNFα concentration in sera of mice bearing immunogenic tumors (3LL and MC57) was increased, whereas TNFα concentrations in sera of MC38 and B16F10 tumor-bearing mice was not (Fig. 6M). Furthermore, an in silico pathway analysis could predict that at gene/protein level, most of the elevated cytokines in sera or TCMs of tumor-bearing mice can potentially influence the expression of TNFα (Fig. 6N). Taken together, these results indicated that TNFα secreted by T cells activates HSPCs in tumor-bearing mice.

To study the function of IL6 and TNFα in vivo, we transplanted MC57 tumor fragments into IL6^−/− or TNFR1/2^−/− mice. Similarly, to BL/6 mice, IL6-deficient tumor-bearing mice had elevated numbers of LSKs and CMPs in the bone marrow (Supplementary Fig S4A). In addition, HSPCs from tumor mice formed more colonies in vitro (Supplementary Fig S4B).

In line with results in other tumor models (37), transplanted MC57 tumors did only grow for up to two weeks and were then rejected in TNFR1/2^−/− mice (Fig. 7A). Therefore, we analyzed hematopoiesis in the bone marrow already 14 days after tumor transplantation, at a time point where tumors in BL/6 and TNFR1/2^−/− mice were very small. Analysis of bone marrow revealed a slight increase in LSKs and CMPs in BL/6 tumor-bearing mice compared with naïve controls. In contrast, TNFR1/2^−/− tumor-bearing mice had no increase in these cell populations (Fig. 7B and C). In addition, MDSCs in bone marrow of tumor-bearing mice were increased in BL/6 but not in TNFR1/2^−/− mice (Fig. 7D).

To further analyze the role of TNFα in the regulation of myelopoiesis in a second tumor model, we transplanted 3LL-Lewis lung carcinoma cells to TNFR1/2^−/− and BL/6 mice. TNFα was significantly increased in the sera and in TCM of CD4⁺ T cells from 3LL tumor-bearing mice (Fig. 7E). Loss of function of TNFα in TNFR1/2-deficient mice normalized hematopoiesis in the bone marrow with comparable numbers of LSKs, CMPs, and GMPs in naïve and tumor-bearing mice (Fig. 7F). Similarly, colony formation of Lin⁻ cells from 3LL tumor-bearing TNFR1/2^−/− mice was comparable with naïve controls (Fig. 7G). A subdifferentiation of the LSK compartment revealed that the number of the primitive HSC subsets (LT-HSC and ST-HSC) remained unchanged, whereas the numbers of MPP1 and MPP2 increased significantly in 3LL tumor-bearing BL/6 mice. Importantly, MPP1 and MPP2 cell numbers in tumor-bearing TNFR1/2^−/− mice were comparable with naïve mice, indicating that the increase in MPP1 and 2 cells in BL/6 tumor-bearing mice is dependent on TNFα signaling. The number of MPP3, which comprises mainly lymphoid progenitors remained unchanged in BL/6 tumor-bearing mice and even dropped in TNFR1/2^−/− tumor-bearing mice (Fig. 7H). Numbers of Mo- and Gr-MDSCs in spleen were similarly reduced in TNFR1/2^−/− tumor-bearing mice as compared with BL/6 tumor-bearing mice (Fig. 7I). Furthermore, TNFα increased cell cycling of LSKs in tumor-bearing BL/6 mice, as indicated by a higher frequency of LSKs in S₁ phase, but not in TNFR1/2^−/− mice (Fig. 7J). Similarly, neutralization of TNFα in 3LL tumor-bearing mice by treatment with a mAb blocked the increase in LSKs, CMPs, and MDSCs (Fig. 7K–M). Moreover, TNFα depletion significantly reduced tumor growth (Fig. 7N). These results confirm that TNFα increases myelopoiesis and the accumulation of MDSCs in tumor-bearing mice.

Escape from immunosurveillance is a hallmark of cancer development (38). Thereby, tumor cells adopt strategies to overcome destruction by tumor antigen–specific effector cells. For example, cancer cells generate an immunosuppressive microenvironment in the tumor by producing immunosuppressive factors such as PD-L1, FasL, IL10 or TGFβ that directly inhibit the activity of antitumoral effector cells or by recruiting immunosuppressive cells such as Tregs and MDSCs (39). Importantly, the level of immunosuppression is a negative prognostic factor in patients with cancer (4, 40–43). MDSCs are one important cell population that is recruited to the tumor microenvironment and induces immunosuppression (5, 7, 44). Unlike lymphocytes, myeloid cells including MDSCs do not have the capacity for clonal expansion and, in addition, they have a relatively short half-life in vivo (45). In cancer, a high activity of reactive oxygen species (ROS) in MDSCs increases apoptosis and even reduces the life span of MDSCs (9). Thus, MDSCs need to be replenished continuously from hematopoietic precursors in the bone marrow. Some tumor cells produce hematopoietic cytokines such as GM-CSF, G-CSF, and IL6 that increase myelopoiesis in the bone marrow and the production of MDSCs (6–8, 46). However, a majority of tumor cells does not produce hematopoietic growth factors and the mechanisms regulating hematopoiesis in these tumor types remain unknown (8). We now show that an activated adaptive immune system regulates hematopoiesis in different experimental tumor models. T cell–secreted TNFα induced emergency hematopoiesis by increasing cell cycling activity of LT- and ST-HSC and myeloid progenitors. Interestingly, analysis of HSC subsets revealed increased numbers in myeloid-skewed MPP1 and MPP2 subsets but not lymphoid-skewed MPP3s. This preferential differentiation to the myeloid lineage was confirmed by increased numbers of myeloid progenitors (CMPs and/or GMPs) with normal numbers of CLPs.

T lymphocytes control hematopoiesis through various mechanisms. In steady state, effector CD4⁺ T cells in the bone marrow regulate myelopoiesis and ensure terminal differentiation of myeloid cells by secreting IL6, IL3, and GM-CSF (47). During inflammation, cytotoxic CD8⁺ T cells secrete IFNγ that acts on HSCs and impairs their self-renewal by dephosphorylating STAT5 (48). In addition, activated T lymphocytes produce hematopoietic cytokines, such as colony-stimulating factors, IL6 and IL17 that induce myeloid differentiation and amplify granulocyte production (47). Coculture experiments and CD4⁺ and CD8⁺ T-cell depletion experiments revealed that the factors regulating hematopoiesis in our tumor models are mainly derived from CD4⁺ T cells. We found that many important cytokines that are directly or indirectly involved in the regulation of the hematopoiesis such as TNFα and IL6 are increased in the sera and in CD4⁺ TCM from tumor-bearing mice.

We excluded a major role of CD4⁺ T cell–secreted IL6 in the regulation of myelopoiesis in vitro by adding IL6-neutralizing antibodies in colony-forming assays and in vivo by transplanting tumors in IL6-deficient mice. TNFα was the cytokine with the largest increase in CD4⁺ TCM and neutralization of TNFα prevented the increase in colony-forming capacity of BL/6 LSKs induced by the sera of tumor-bearing mice. Although TNFα is produced by CD8⁺ and CD4⁺ T cells, in vivo depletion of each cell population individually suggests that the main effect on hematopoiesis is mediated by TNFα secreted by CD4⁺ T cells. In contrast, CD4⁺ and CD8⁺ T-cell depletion similarly reduced MDSC numbers in tumor-bearing mice, suggesting that CD8⁺ T-cell influence the differentiation to MDSCs by additional mechanisms. Importantly, transplantation of MC57 and 3LL tumors to TNFR-deficient mice indicated that TNFα leads to the activation of hematopoiesis, myeloid skewing, and to the increase in MDSCs observed in tumor-bearing mice. This central role of a single cytokine in the regulation of the hematopoiesis in tumor-bearing mice was somewhat surprising, because at least 41 of the 48 cytokines studied were detected at higher or lower levels in the sera of tumor-bearing mice versus naïve mice. However, an in silico pathway analysis suggested that TNFα interacts with and possibly regulates most of the other molecules analyzed.

TNFα signaling through its receptors (TNFR) TNFR-I (p55) and TNFR-II (p75) activates NF-κB and other signaling pathways that increases cell survival, activation, and proliferation (49). The soluble form of TNFα (sTNFα) mainly triggers TNFR-I, whereas the transmembranous form (tmTNFα) preferentially activates TNFR-II (50) with distinct biological functions. It has been documented that mainly sTNFα drives the differentiation and accumulation of MDSCs in a MCA tumor model (14). The function of TNFα in the regulation of hematopoiesis has been analyzed in different models with partially contradictory results: TNFα has been shown to suppress or increase the colony formation capacity of HSCs and their in vivo reconstitution capacity, probably depending on the dose and length of exposure studied (51, 52). Prolonged and excessive TNFα has been associated with myelodysplastic syndromes (53). Furthermore, mice deficient of the p55 TNFR 1α (TNFRSF1α^−/−) have increased numbers of functionally impaired HSPCs as indicated by a reduced self-renewal capacity (54). Tumor development and consecutive activation of the adaptive immune system lead to a continuous production of TNFα. The transplantable tumor models have the limitation that tumors develop fast and long-term effects on hematopoiesis cannot be studied. However, the development of MCA-induced sarcoma and of lung adenocarcinoma in the KP model takes several months and therefore mimic the physiologic development of a cancer and of the antitumoral immune response more closely. Importantly, we observed a similar increase in HSPCs with myeloid skewing independent of the growth kinetics of the tumor. Experiments using TNFR1/2^−/− confirmed that TNFα is an important cytokine in activating myelopoiesis and, thereby, contributes to an increase in MDSC numbers. In vitro experiments with TCM suggest that the sTNFα is the main driver of the expansion of HSCs. However, TNFα does not only increase MDSC numbers by regulating HSPCs. TNFα signaling directly enhances the survival of MDSCs through cellular FLICE-inhibitory protein (c-FLIP)-mediated inhibition of caspase-8 (37). Together, these mechanisms contribute to the tumor resistance of TNFα-deficient mice (55).

The fact that the activated antitumoral immune response by itself leads to an increase in myelopoiesis and MDSCs and, thereby, to immunosuppression indicates a physiologically important regulatory system. Several comparable regulatory circuits have been described. The effector cytokine IFNγ secreted by activated T cells induces the upregulation of the T-cell–inhibitory ligand PD-L1 (56). Ligation of CD27 by CD70 expressed on activated immune cells induces the expansion of Tregs and leads to an impaired tumor immunosurveillance (57). IL5 and IL13 cytokines produced by Th2 cells enhance type II macrophage differentiation (58). Therefore, tumors escape immunosurveillance by regulatory circuits that developed to prevent immunopathology. Defining and blocking these mechanisms led and may lead to promising therapeutic strategies to treat cancer.

No potential conflicts of interest were disclosed.

Conception and design: M.F. Al Sayed, C. Riether, A.F. Ochsenbein

Development of methodology: M.F. Al Sayed, C. Riether

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M.F. Al Sayed, M.A. Amrein, E.D. Bührer, R. Radpour

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M.F. Al Sayed, M.A. Amrein, R. Radpour, A.F. Ochsenbein

Writing, review, and/or revision of the manuscript: M.F. Al Sayed, M.A. Amrein, C. Riether, A.F. Ochsenbein

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M.F. Al Sayed, M.A. Amrein, A.-L. Huguenin

Study supervision: M.F. Al Sayed, A.F. Ochsenbein

This work was supported by the Swiss National Science Foundation, the Swiss Cancer League, and the Werner und Hedy Berger-Janser-Stiftung.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.