MC-38GFP murine colon carcinoma cell line (18), LLC1.1 Lewis lung carcinoma, 3LL lung carcinoma, and B16-BL6 were grown in DMEM high glucose medium (Life Technologies) containing 10% fetal bovine serum (FBS) as described previously (19).

LysMCre/Tgfbr2^fl/fl mouse in a C57BL/6J background was obtained from M. Heikewälder (University of Zurich). The mouse strain ROSA-STOP^fl/fl-dTomato was obtained from The Jackson Laboratory. All mice were in the C57BL76J background. Both males and female animals in the age of 7-10 weeks of age were used in all experiments. All animal experiments were performed according to the guidelines of the Swiss Animal Protection Law, and approved by the Veterinary Office of Kanton Zurich.

Spontaneous lung metastasis. Lewis lung carcinoma cells LLC1.1 (300,000 cells) were subcutaneously (s.c.) injected into a right flank of a mouse. The primary tumor was removed after 18 days and the extent of lung metastasis was evaluated two weeks after tumor removal. Experimental liver metastasis. MC-38GFP colon cancer cells (300,000 cells) were injected into the spleen of an anesthetized mouse. Splenectomy was performed three minutes after the tumor cell injection. Mice were terminated on day 28 and liver metastasis was evaluated. Experimental lung metastasis. MC-38GFP (300,000 cells) were intravenously injected and mice were terminated 28 days after the injection. 3LL and B16-BL6 cells (150,000 cells) were intravenously injected and lungs analyzed on day 14 (19). Surface metastasis were counted in perfused lung and liver tissues.

Tissues paraffin sections (3 μm) were stained with hematoxylin/eosin or the anti-GFP antibody (Fitzgerald Industries Int.) Ki67 (NeoMarkers) and cleaved caspase-3 (BD). Quantification of tumor cells in seven sections of lungs separated by 20 μm was performed. Staining was performed on a NEXES immunohistochemistry robot using an IVIEW DAB Detection Kit (Ventana Instruments, Switzerland). Images were digitalized on Zeiss MiraxMidi Slide Scanner and analyzed with Panoramic Viewer (3DHISTECH). For cryosections, tissues were fixed in 3% paraformaldehyde for 1 hour at 4°C, incubated in 20% sucrose overnight at 4°C, embedded in an OCT compound and snap-frozen in a bath containing methylbutane and dry ice. Cryosections (10 μm) were blocked with phosphate buffer saline (PBS) containing 5% FCS and 5% rat serum, and incubated with AlexaFluor 647-coupled rat-anti-F4/80 antibody (Clone BM8, Biolegend). Slides were counterstained with DAPI and embedded in Prolong and evaluated on Leica Thunder Microscope.

Mice were perfused with PBS and dissected tissues were minced and digested with Collagenase A and D (1 mg/mL, both from Roche), in 2 mL of RPMI medium containing 3% BSA for 1 hour at 37°C. The cell suspension was prepared by passing the homogenate five times through 18G needle followed by a 100μm cell strainer. In case of liver, the cell suspension was centrifuged at 500 rpm for one minute to separate the parenchyma. After centrifugation of the single cell suspension, erythrocytes lysed with ACK buffer and cells were passed through a 70μm cell strainer. The pellet was resuspended in a FACS buffer (PBS/2% FBS/5mM EDTA) and stained with Zombie Fixable Viability Kit (Biolegend), followed by incubation with anti-mouse CD16/32 mAb for 10 min on ice. Single cell suspensions were stained with the following antibodies: anti-CD45 (clone 30-F11), anti-CD11b (clone M1/70), anti-F4/80 (clone BM8), anti-Ly6G (clone 1A8), anti-Ly6C (clone HK1.4), anti-MHCII (clone M5/114.15.2), anti-CD31 (clone 390), CD4 (clone GK1.5), anti-CD8 (clone 53-6.7) and anti-PD1 (clone 29F.1A12) (all from Biolegend) for 30 min on ice. Data was acquired on a FACSCanto II (BD Biosciences), during cell quantification together with CountBright absolute counting beads (Life Technologies) and analyzed using Flow Jo software v. 7.6.5 (Tree Star). Samples for CD8 cell activation assessment were fixed and permeabilized using FoxP3/Transcription Factor Staining Buffer Set (eBioscience). Subsequently samples were stained for intra-cellular markers with the following antibodies: anti-Granzyme B (clone GB11) and anti-Perforin (clone S16009A) from Biolegend. CD8 cell activation was analyzed using the Cytek Aurora 5L – Spectral Analyzer from the Cytometry Facility at the University of Zurich.

Tissues were processed the same way as described for flow cytometry analysis. Individual cell populations were sorted using a BD Aria III in FACS buffer at 4°C using a 100 μm nozzle. From at least 100’000 sorted cells RNA was isolated using the RNA Easy Plus Mini kit (Qiagen) and cDNA was prepared using 250 ng of RNA using Omniscript RT kit (Qiagen) according to the manufacturer’s instructions. Real-time PCR was performed with KAPA Sybr FAST Universal (Sigma) using intron-spanning primers ( Supplementary Table S1 ) and analyzed in a CFX96 Touch Real-Time PCR Detection System (Biorad). GAPDH was used as an internal control.

Red fluorescence beads (Sigma) were diluted 1:20 in PBS and 100 μl was intravenously injected through a tail-vein per mouse. Mice were terminated after 9 hours, blood and organs were collected and processed as described in flow cytometry analysis.

Tumor bearing mice (subcutaneous LLC1.1 tumor) were anesthetized, the lungs perfused with PBS and through heart perfused with pre-warmed 5 μM CellROX^® Oxidative Stress Green Reagent (Invitrogen) and the vessels clipped. After 30 min of incubation, the lungs were again perfused with PBS, digested with collagenase and single cell suspension stained with anti-Ly6G antibody and the ROS signal was quantified by flow cytometry.

MC-38-luciferase expressing cells (300,000) were injected through spleen in mice and at indicated time points, mice were anesthetized, subcutaneously injected with D-luciferin firefly (150 μg/g mouse) and after 8 min terminated. Mice were perfused with PBS and the luciferase activity in the liver was determined using IVIS imaging. After the intravenous injection of 300,000 MC-38GFP cells mice were terminated at indicated time points; mice perfused with PBS and lungs processed for histological analysis.

Depletion of CD8⁺ T cells in a lung metastasis model using LLC1.1 cells was performed with 10 μg anti-CD8 antibody (clone 2.43; BioXCell) i.v. injected one day prior to primary tumor removal. Mice received another 15 μg of the antibody by i.v. injection two days after the tumor removal. Control mice received the isotype control antibody (clone 2A3, BioXCell) in parallel. In the liver metastasis model, mice received 10 μg Ab injection 1 day before tumor cell injection and the second treatment with 15 μg Ab two days after tumor cell injection.

Neutrophil depletion was achieved by intraperitoneal injection (i.p.) of anti-Ly6G antibody (clone 1A8; BioXCell). Initial injection of 500 μg/mouse was performed 24 hours prior to intrasplenic injection of MC-38GFP cells. Mice were treated three times weekly with 100 μg i.p. until the termination at day 27. Control animals were injected with isotype-control Ab (clone 2A3, BioXCell) in parallel.

Snap-frozen tissues were homogenized using Polytron^® in lysis buffer (FACS buffer containing 1x complete EDTA free proteinase inhibitor cocktail from Roche). After 15 min centrifugation at 14’000 rpm at 4°C, was determined the protein concentration in the supernatant was determined with BCA Protein Assay (Thermo Fischer). Cytokine concentrations were measured in a total of 200-500 μg of tissue lysate using the ProcartaPlex 23 Plex and the LegendPlex Mouse Proinflammatory Chemokine panel (Thermo Fischer) following the manufacturer’s instructions. TGFβ1 was determined using the TGFβ Platinum Elisa (Thermo Fisher), where samples were activated with 1N HCl for 10 min at RT in a ratio of sample to HCl 5:1. The reaction was neutralized using 1 part of 1N NaOH/0.5M HEPES. Sorted tdT⁺ Ly6C^hi monocytes (20’000 cells/well) from livers of mice at day 14-post-intrasplenic injection of MC-38GFP cells were stimulated overnight (14 hrs) with PMA (20 ng/ml), ionomycin (1μg/ml) and LPS (1 μg/ml). Cytokine production was detected in the supernatant after 14 h stimulation as described above.

Myeloid cells are the dominant population of immune cells associated with metastatic progression (20–22); and TGFβ is an immunosuppressive cytokine family that promotes metastases through modulations of the tumor microenvironment (7). To assess the role of Tgfβ-signaling in myeloid cells during lung and liver metastasis, we generated a mouse with myeloid cell-specific deletion of TGFβ receptor II (LysMCre⁺/Tgfbr2^fl/fl mouse, hereafter TR2^myKO) by crossing the Tgfbr2^fl/fl mouse with a mouse expressing Cre under the control of the Lysozyme promoter (LysMCre). LysMCre^neg/Tgfbr2^fl/fl (hereafter Ctrl) served as a control.

First, we tested the role of myeloid cell-derived Tgfβ-signaling during liver metastasis. We used an experimental liver metastasis model, where MC-38 murine colon carcinoma cells stably expressing GFP (MC-38GFP) were injected into the spleen. Significant reduction of metastatic foci and liver-body mass ratio was observed in TR2^myKO mice when compared to Ctrl mice ( Figure 1A ). The size of liver metastatic lesions was also significantly reduced in TR2^myKO mice ( Figure 1B ). To test whether myeloid cells affect tumor cell seeding to the liver in a Tgfβ-dependent manner, we intrasplenically injected MC-38-luciferase expressing cells and analyzed livers 9 and 24 hours post-tumor cell-injection using bioluminescence imaging ( Figure 1C ). No difference in tumor cell detection in the liver was observed between TR2^myKO and Ctrl mice, indicating that Tgfβ-signaling pathway in myeloid cells is not affecting initial tissue colonization.

Second, we tested whether Tgfβ-signaling deficiency in myeloid cells affects lung metastasis. Myeloid cells were previously shown to promote metastatic seeding to the lungs (22–24). Using three different cell lines, MC-38, Lewis lung carcinoma 3LL and B16-Bl6 melanoma in experimental lung metastasis models, we observed reduced metastasis in TR2^myKO mice ( Figure 1D and Supplementary Figures 1A, B ). To assess the effect of myeloid Tgfβ-deficiency on lung metastasis, we analyzed tissues sections two, four and seven days post-tumor cell injection. While the number of single tumor cells was similar in TR2^myeloKO and Ctrl mice at every time point, we observed a significant reduction of small clusters (2-10 tumor cells) and large cluster (more than 10 tumor cells) at day 4 and day 7 post-tumor cell challenge in TR2^myKO mice ( Figure 1E ). These findings indicate that the metastatic outgrowth in TR2^myKO mice has been greatly hampered (2-10 cells foci) or impaired (>10 cells foci). Next, we used the spontaneous lung metastasis model, where mice were subcutaneously (s.c.) injected with Lewis lung carcinoma (LLC1.1) cells. At day 18 the primary tumor was surgically removed and after another 14 days lung metastasis was analyzed. While no difference in the primary tumor growth between TR2^myKO and Ctrl mice was observed ( Supplementary Figure 1C ), a significant reduction of lung metastasis was detected in TR2^myKO mice ( Figure 1F ).

To understand the role of myeloid cell Tgfβ-signaling during liver metastasis, we analyzed these cells in metastatic livers at two different time points after tumor cell injection: 14 days – microscopic metastasis, 28 days – macroscopic metastasis, and compared the changes to naïve mice. We observed increased numbers of inflammatory monocytes (Ly6C^hi) and neutrophils (Ly6G⁺) in Ctrl mice, which were significantly reduced in TR2^myKO at day 28 ( Figure 2A ). There were no changes in lymphoid CD4⁺ or CD8⁺ cells at any time point. Next, we analyzed cytokine levels in metastatic livers ( Figure 2B ). Increased levels of Ccl2, Ccl7, Cxcl1, and Cxcl10 chemokines both at day 14 and at day 28 in Ctrl mice correlated well with respective increased cytokine levels ( Figure 2C ). Both lower levels of cytokines and reduced myeloid cell number were detected in TR2^myKO mice. Increased amounts of Tgfβ1 cytokine were detected in metastatic livers at day 14, irrespective of mouse genotype, which decreased to levels observed in naïve mice at day 28 ( Figure 2D ).

Next, we aimed to detect myeloid cells in liver metastatic lesions. We prepared a mouse by breeding of LysMCre⁺/Tgfbr2^fl/fl strain into a transgenic mouse, Ai14, carrying loxP-flanked STOP cassette followed by tdTomato inserted into the ROSA26 locus. We analyzed myeloid cells from naïve LysMCre⁺/Tgfbr2^fl/fl /tdT⁺ mice (TR2^myKO/tdT) and LysMCre/Tgfbr2^wt /tdT⁺, wild-type allele (Ctrl/tdT) mice, and detected over 90% tdT⁺ Ly6G⁺ and over 30% tdT⁺ of Ly6C^hi cells in liver, lungs and spleen ( Supplementary Figure 2 ). When MC-38GFP tumor cells were injected intrasplenically into TR2^myKO/tdT and Ctrl/tdT mice, we observed increased numbers of both Ly6C^hi and Ly6G⁺ cells in Ctrl/tdT mice by flow cytometry, while in TR2^myKO/tdT mice these myeloid cells were hardly present in the metastatic lesions ( Figure 3A ). These results confirm our previous data ( Figure 2A ). The histological analysis of liver sections at day 14 and 28 days post-tumor cell injection substantiated the finding of a pronounced reduction of tdT-positive cells within metastatic foci in TR2^myKO/tdT when compared to Ctrl/tdT mice ( Figure 3B ). Furthermore, large proportion of tdT-positive cells were positive for an F4/80 antigen ( Figure 3C ), indicating differentiation of myeloid-derived cells into macrophages. Histological analysis of metastatic livers for proliferating cells (Ki67⁺ cells) and for apoptotic cells (cleave-caspase3⁺ cells) revealed no obvious differences between Ctrl and TR2^myKO mice ( Supplementary Figure 3 ). These data show a reduced presence of Tgfβ-responsive myeloid cells inside the growing metastatic foci, which correlated with decreased amounts of chemokines in the metastatic livers in TR2^myKO mice.

To assess how Tgfβ-responsive myeloid cells promote liver metastasis, we first analyzed myeloid cells from metastatic livers of TR2^myKO and Ctrl mice. First, the phagocytic activity of myeloid-derived (CD11b⁺) cells was tested. We analyzed Ly6C⁺ monocytes and Ly6G^lo/Ly6C^- macrophages isolated from livers of mice intrasplenically injected with MC-38 cells at day 14, during early metastasis. No differences in phagocytic capacity of monocytic cells were observed between the respective mouse genotypes ( Figure 4A ). A recent study using a murine breast cancer model with 4T1 tumor cells in LysMCre⁺/Tgfbr2^fl/fl mice showed that CD8⁺ T cells control lung metastasis (16). Thus, we depleted CD8⁺ T cells to test whether myeloid cells modulate CD8⁺ T cells and thereby promote liver metastasis. While efficient depletion of CD8⁺ T cells in the blood liver and lungs has been achieved ( Supplementary Figure 4A ), no effect on liver metastasis has been observed ( Figure 4B ). This finding points to a different mechanism of Tgfβ-stimulated pro-metastatic myeloid cell involvement during liver metastases. To determine the subpopulation of myeloid cells involved in liver metastasis, we depleted Ly6G⁺ cells during experimental liver metastasis. Depletion of neutrophils showed no difference in liver metastasis neither in Ctrl nor in TR2^myKO mice ( Figure 4C ), suggesting the involvement of myeloid-derived monocytic cells.

The altered polarization of monocytes/macrophages and neutrophils towards M2 and N2 phenotype, respectively, was previously associated with cancer progression (25, 26). Therefore, we sorted Ly6C^hi monocytes and Ly6G⁺ neutrophils from metastatic livers of TR2^myKO and Ctrl mice and analyzed the gene expression associated with M1 and M2 polarization (27) and N1 and N2 polarization (28), respectively. We observed an increased expression in seven out of eight genes in monocytic cells from TR2^myKO mice, which were mostly M1 genes ( Figure 4D ). Specifically, increased expression of ICAM1, TNFα and SOCS3, and decreased expression of M2 genes, CCL2 and Arg1 was detected in Ly6C^hi cells from TR2^myKO mice ( Figure 4D ). However, we observed no significant changes in polarization of Ly6G⁺ cells, while only the ICAM1 was increased in cells from TR2^myKO mice. To confirm the altered polarization of monocytes, we sorted Ly6C^hi cells from metastatic livers and measured cytokine production in vitro ( Figure 4E ). Ly6C^hi cells from TR2^myKO mice showed increased production of Il-1β, and partially increased Tnfα levels when compared to monocytes from Ctrl mice. Reduced levels of several cytokines, such as Il-6, Ccl2, Ccl3, Ccl5, Cxcl1 and Cxcl10, but also Il-10 were detected in Ly6C^hi cells from TR2^myKO mice ( Figure 4E and Supplementary Figure 4B ). Changes in cytokine production in Ly6C^hi cells supports the active role of Tgfβ-signaling in modulation of myeloid-derived cells that facilitates liver metastasis.

We studied how the absence of Tgfb-signaling in myeloid cells reduces lung metastasis. The cellular composition of metastatic lungs in the LLC1.1 spontaneous metastatic model was analyzed. We selected day 21, which is 5 days after the primary tumor removal, as an early metastatic time point; and day 32, as a late metastatic phase. No differences in the presence of Ly6C^hi monocytes, Ly6G⁺ granulocytes and CD4⁺ T cells were detected between TR2^myKO and Ctrl mice at any time point ( Figure 5A ). However, we observed increased numbers of CD8⁺ T cells in the lungs of TR2^myKO mice at day 32 ( Figure 5A ). Of note, reduced numbers of F4/80⁺ macrophages was observed in the lungs of TR2^myKO mice at day 21, but not at day 32, when compared to control mice ( Supplementary Figure 5A ). Next, we analyzed cytokine levels in metastatic lungs at day 21 ( Figure 5B ). There was an overall decrease in cytokines in TR2^myKO mice when compared to Ctrl mice, specifically a significant reduction of Ifnγ, Il-4, Il-5,Ccl2 and Ccl11 was observed. Since tumor-associated myeloid cells can inhibit immune cell functions (20), the observed decrease in Il-4 and Il-5 together with a tendency of Il-10 in TR2^myKO mice suggests a possible reduced polarization of T cells. The reduced Ccl2 levels detected in TR2^myKO mice corresponds well with reduced macrophages in the lungs. Finally, we measured the amounts of Tgfβ1 cytokine in the lungs of tumor-bearing mice and compared them with non-injected (naïve) mice ( Figure 5C ). We observed an increase in Tgfβ1 levels in the lungs of tumor-bearing mice at day 14, while at day 28 there was no difference when compared to the lungs of naïve mice. When we analyzed metastatic lungs from TR2^myKO/tdT and Ctrl/tdT mice at day 32 days after LLC1.1 s.c. injection, we observed no changes in the presence of tdT-positive Ly6C^hi and Ly6G⁺ cells between mouse phenotypes as determined by flow cytometry ( Supplementary Figure 5B ).

Next, we tested whether attenuated lung metastasis is associated with altered polarization of Tgfβ-signaling deficient myeloid cells, when no major changes in myeloid cell presence in lungs of TR2^myKO/tdT and Ctrl/tdT mice were detected ( Figure 5A ). First, the phagocytic activity of myeloid-derived (CD11b⁺) cells was tested. We analyzed Ly6C⁺ monocytes and Ly6G^lo/Ly6C^- macrophages isolated from lungs of mice injected with LLC1.1 cells at day 21, during early metastasis. No changes in uptake of beads neither in monocytic cells nor myeloid-derived macrophages isolated from TR2^myKO and Ctrl mice could be detected ( Figure 6A ). Second, we tested the production of reactive oxygen species (ROS) by myeloid-derived Ly6G⁺ cells, which was similar in both mouse genotypes ( Figure 6B ). These data indicate no functional changes in Tgfbr2-deficient myeloid cells in metastatic sites. TGFβ signaling is a critical mediator of immune cell polarization (7). Therefore, we sorted Ly6C^hi monocytes and Ly6G⁺ neutrophils from the lungs at day 21, early metastasis, and analyzed their polarization status on a transcriptional level. However, we detected no significant changes in expression of genes linked to respective polarization patterns of myeloid cells ( Figure 6C ). Since we observed enhanced infiltration of lungs metastasis with CD8⁺ T cells in TR2^myKO mice ( Figure 5A ), we tested their role in control of lung metastasis. Depletion of CD8⁺ T cells in a LLC1.1 model 24 h before the primary tumor removal and during the metastatic phase resulted in increased lung metastasis in TR2^myKO mice to similar levels as observed in Ctrl mice ( Figure 6D ), indicating that Tgfβ-signaling-deficient myeloid cells control lung metastasis through modulation of CD8⁺ T cells. The analysis of CD8⁺ T cells from metastatic lungs ( Supplementary Figure 6 ) showed a significantly increased levels of perforin⁺ CD8⁺ T cells in TR2^myKO mice when compared to Ctrl mice ( Figure 6E ). However, no major differences in granzyme B expression, nor alteration in PD1 expression has been observed. These data provide evidence that the absence of Tgfβ-signaling in myeloid cells results in increased levels of functional CD8⁺ T cells, which control metastatic outgrowth in the lungs.