Tumor-educated Tregs drive organ-specific metastasis in breast cancer by impairing NK cells in the lymph node niche

Breast cancer is accompanied by systemic immunosuppression, which facilitates metastasis formation, but how this shapes organotropism of metastasis is poorly understood. Here, we investigate the impact of mammary tumorigenesis on regulatory T cells (T regs ) in distant organs and how this affects multi-organ metastatic disease. Using a preclinical mouse mammary tumor model that recapitulates human metastatic breast cancer, we observe systemic accumulation of activated, highly immunosuppressive T regs during primary tumor growth. Tumor-educated T regs show tissue-specific transcriptional rewiring in response to mammary tumorigenesis. This has functional consequences for organotropism of metastasis, as T reg depletion reduces metastasis to tumor-draining lymph nodes, but not to lungs. Mechanistically, we find that T regs control natural killer (NK) cell activation in lymph nodes, thereby facilitating lymph node metastasis. In line, an increased T reg /NK cell ratio is observed in sentinel lymph nodes of breast cancer patients compared with healthy controls. This study highlights that immune regulation of metastatic disease is highly organ dependent.

Here we describe how mammary tumors systemically rewire T, and how this affects metastatic disease to different organs. To achieve this, we utilized models that allow for interrogation of tissue-specific metastasis, i.e. the transgenic K14cre;Cdh1F/F;Trp53F/F (KEP) mouse model of invasive mammary tumorigenesis (), and the KEP-based mastectomy model for spontaneous multi-organ metastatic disease (). We observed systemic accumulation of activated, highly immunosuppressive Tduring primary tumor growth. These Tshowed striking tissue-specific transcriptional rewiring in response to mammary tumorigenesis, and elicited a tissue-specific effect on metastasis formation, as neo-adjuvant depletion of Treduced cancer spread to axillary (Ax) LNs, but not to the lungs. Mechanistically, we demonstrate that Tpromote LN metastasis formation through inhibition of NK cells in the LN niche. These findings add another mechanism to the emerging body of literature that immune regulation of metastatic disease is highly organ dependent, warranting a more personalized approach in the fight against metastatic disease.

Somatic inactivation of E-cadherin and p53 in mice leads to metastatic lobular mammary carcinoma through induction of anoikis resistance and angiogenesis.

Despite these intriguing clinical observations, and the attention that tumor-associated Thave received in the context of breast cancer in recent years (), the lack of preclinical models that closely recapitulate human multi-organ metastatic disease has limited our understanding of the importance of Tin cancer spread to different distant organs (). Preclinical studies performed with mouse models based on orthotopic inoculation of breast cancer cell lines have shown that ablation of Tcan attenuate primary tumor growth and subsequent metastasis formation to the lungs (). However, research on Tin the context of cancer is mostly focused on their role in the microenvironment of primary tumors or metastases. The systemic impact of primary tumors on Tin distant organs, and their functional significance for metastasis formation in different tissue contexts, has remained largely unclear. Additionally, the role of Tin hallmarks of metastatic disease such as systemic immunosuppression and the development of a pre-metastatic niche is understudied (), and therefore remains elusive.

An important cell type involved in immunosuppression in cancer is the CD4FOXP3regulatory T cell (T) (). In breast cancer, immunosuppressive Tdensely populate human tumors, and high levels of intratumoral Tcorrelate with high tumor grade and poor survival (). Intriguingly, clinical data suggest that primary breast tumors affect Tbeyond the tumor microenvironment. Tin peripheral blood have been reported to be increased in breast cancer patients (), and their responsiveness to cytokine stimulation is predictive of breast cancer relapse (). In addition, recent studies have shown that Taccumulate in sentinel lymph nodes (LNs) of breast cancer patients, which correlates with cancer spread to these LNs (), suggesting a potential role for Tin modulating metastasis to tumor-draining LNs.

Prevalence of regulatory T cells is increased in peripheral blood and tumor microenvironment of patients with pancreas or breast adenocarcinoma.

The main cause of breast cancer-related mortality is metastatic disease. Over the past decades, breast cancer survival has improved through detection and intervention in early stages of breast cancer, but preventing and treating metastasis remains an unmet clinical need (). Disseminated cancer cells progress through a multistep cascade, which involves complex interactions between cancer and host cells, including immune cells (). The immune system plays a dual role in metastasis formation. While properly activated cytotoxic immune cells are equipped to control metastasis, tumor-induced immunosuppressive immune cells exploit a diversity of mechanisms to promote metastasis (). Emerging data indicate that tissue tropism of metastasis may be influenced by the immune contexture in distant organs, suggesting an additional layer of complexity in metastasis formation (). However, how immunosuppressive mechanisms differ per metastatic site, and how this shapes tissue tropism of metastasis, is poorly understood.

Finally, we validated our preclinical findings on Tand NK cell interactions in the LN niche of breast cancer patients. To do so, we analyzed the accumulation of T(CD4CD25FOXP3) and NK cells (CD56CD16and CD56CD16) in tumor-free and tumor-positive sentinel LNs of breast cancer patients (BrCa SLN/SLN), and in Ax LNs from healthy controls (HLNs), using a previously described flow cytometry dataset (). In line with previous analyses of this unique dataset () and consistent with our preclinical data ( Figure 1 C), Tlevels are significantly elevated in BrCa SLNs compared with HLNs ( Figure 6 C). We also observed a statistically significant reduction of CD56CD16, but not CD56CD16NK cells, in BrCa SLN, and a similar trend in BrCa SLN Figures 6 D and 6E). Notably, in particular CD56CD16have been described to have cytotoxic activity (). Combined, this shifts the T/NK cell ratio strongly toward Tin both tumor-free and tumor-positive BrCa SLNs compared with HLNs ( Figure 6 F). Despite the low number of BrCa SLNsamples, we also observed a non-significant trend of a higher T/NK cell ratio in SLNversus SLNsamples. A rise in Tin conjunction with a reduction of potentially cytotoxic NK cells in the SLN niche is in accordance with our preclinical finding that LN NK cells have reduced expression of CD107a and the cytotoxic molecule granzyme B under control of Tin tumor-bearing mice ( Figures 5 B–5D).

We next assessed whether the inhibitory effect of Ton Ax TDLN NK cells affects their capacity to control LN metastasis formation. We performed neo-adjuvant co-depletion of Tusing anti-CD25-M2a and NK cells using anti-NK1.1 in the KEP metastasis model. Anti-NK1.1 efficiently depleted NKp46DX5NK cells in the blood of KEP tumor-bearing mice ( Figure S5 C). Strikingly, although depletion of Tsignificantly reduced the incidence of Ax LN metastasis, combined depletion of Tand NK cells completely restored LN metastasis formation ( Figure 6 A ). Anti-NK1.1 alone did not alter LN metastasis incidence and none of the treatments affected the number of lung metastases ( Figure 6 B). Combined, our findings show that tumor-educated Trepress NK cell activation in Ax TDLNs, thereby curbing their anti-metastatic potential, leading to an increased incidence of LN metastasis. This T-mediated immune escape mechanism is specific to the Ax LN, as Tdid not control lung metastasis in this model. Because we did observe some activation of lung NK cells at the transcriptional level in T-depleted versus non-depleted mice ( Figure 5 I), we hypothesized that additional layers of immunosuppression in the lung microenvironment that are independent of Tmay hinder the anti-metastatic potential of lung NK cells. In support of this hypothesis, we found that lung NK cells are mostly terminally differentiated (CD27CD11b) in tumor-free mice, but undergo a partial shift toward a non-cytotoxic immature phenotype (CD27CD11b) in tumor-bearing mice, independent of T Figure S5 D). In contrast to lungs, and in line with previous literature (), Ax TDLN NK cells were found to be mostly in CD27CD11b(immature) and CD27CD11b(cytotoxic) states ( Figure S5 E), highlighting the differences between NK cells in LNs and lung. Importantly, maturation status in Ax TDLN NK cells was not affected in tumor-bearing mice, suggesting this mechanism is specific to lungs, and potentially contributes to observed differences between lung and LN.

(F) Ratio of T reg (% of CD4 + ) versus CD56 low CD16 + (% of CD3 − ) in human HLNs (n = 16), BrCa tumor-negative (n = 7) and tumor-positive (n = 3) SLNs. Data in (B)–(F) show mean ± SEM p values are determined by Kruskal-Wallis test with Dunn's correction for multiple comparisons (C–F), Fisher’s exact test (A). ∗∗ p < 0.01, ∗∗∗ p < 0.001, ∗∗∗∗ p < 0.0001.

(D and E) Frequencies of indicated subset of NK cells of total live cells in human HLNs (n = 16), BrCa tumor-negative (n = 7) and tumor-positive (n = 3) SLNs.

(A) Percentage and number of mice with detectable micro-/macroscopic metastases in Ax TDLNs, in mice receiving neo-adjuvant treatment as indicated (n = 21–31 mice/group, data pooled from two independent experiments).

To identify which molecular pathways are controlled by Tin NK cells in tumor-bearing mice, we performed GSEA analysis on the differentially expressed genes of both lung and Ax TDLN NK cells from T-depleted versus Tnon-depleted mice using the MSigDB Hallmark Gene sets, which represent 50 well-defined biological processes () ( Figures 5 H and S5 B). We found that the depletion of Tinduces the upregulation of molecular pathways related to DNA replication (G2M checkpoint, E2F targets, mitotic spindle) and inflammation (inflammatory response, IFNγ response) in both Ax TDLN and lung NK cells, suggesting a common role of Tin curbing NK cell proliferation and activation. However, we also identified pathways that were uniquely upregulated in either Ax TDLN NK cells (IL6-JAK-STAT3 signaling, IL2-STAT5 signaling) or lung NK cells (IFNα response, TNFα signaling via nuclear factor κB [NF-κB]). Although both Ax TDLN and lung NK cells show signs of activation upon depletion of Tbased on GSEA, we identified 676 genes to be uniquely upregulated in Ax TDLN NK cells of T-depleted versus T-non-depleted mice, compared with 326 in lung NK cells ( Figure 5 I). Interestingly, a subset of genes found specifically upregulated in Ax TDLN NK cells of T-depleted mice encode for proteins with immunomodulatory properties that were not found in lung NK cells of T-depleted animals, including Gzmb, which we had previously identified in our FACS-based analyses of NK cells ( Figures 5 B and 5C). Furthermore, we identified other genes encoding for proteins involved in cytotoxicity (Gzma, Serpinb9b), migration (Ccl4, Ccl8, Ccl22, Cxcr6), co-stimulatory receptors (Icosl, Tnfrsf4), and co-inhibitory receptors (Tigit, Lag3, Ctla4, Klrg1), which are indicative of activated NK cells. In summary, these data show that Tregulate NK cells in a tissue-specific manner, and suggest that tissue context does not only drive Tphenotype but also affects their interactions with target cells such as NK cells.

To further dissect the differential impact of Ton NK cells in the LN and lung niche in vivo, we conducted bulk RNA-seq analysis on FACS-sorted NK cells isolated from T-depleted and T-proficient tumor-bearing mice ( Figure 5 E). Notably, anti-CD25 induces depletion of Tvia antibody-dependent cell-mediated cytotoxicity (ADCC) through engagement of Fc receptors () on innate effector cells, including NK cells (). To confirm that the observed activation of NK cells upon antibody-mediated depletion of Tis independent of their role in ADCC, we now utilized Foxp3mice in which FOXP3cells are efficiently depleted upon injection of diphtheria toxin (DT) ( Figure S5 A). NK cells were obtained from lungs and Ax TDLNs of PBS or DT-treated Foxp3mice bearing transplanted KEP tumors. Gene expression analysis of NK cells from T-depleted versus Tnon-depleted mice identified 1,036 and 646 genes to be differentially expressed in the LNs and lungs respectively, showing that the influence—directly, indirectly, or due to NK cell intrinsic differences—of Ton the NK cell transcriptome is more pronounced in Ax TDLNs than in lungs ( Figures 5 F and 5G, Table S2 ).

To study this as close to the in vivo situation as possible, we assessed the impact of tumor-educated Ton immune cells with potential anti-tumor activity in Ax TDLNs in vivo, instead of using traditional in vitro suppression assays in which cells may lose their functionality imposed by their respective tissue-microenvironment. In vitro suppression assays therefore fail to reproduce the complex interactions that exist in vivo, rendering these assays of limited value for studying metastatic niche-dependent processes. Instead, we depleted Tin mice bearing transplanted KEP tumors and analyzed the phenotype and function of T and NK cells in Ax TDLNs compared with control-treated and naive mice when primary tumors reached a size of 100 mmex vivo. We also analyzed T and NK cells in tumors, blood, and lungs, to gain insights into the tissue-specific impact of tumor-educated Ton these cells. Interestingly, increased expression of the cytotoxic molecule granzyme B by NK cells (CD3, NKp46, DX5) was observed in the Ax TDLNs of tumor-bearing mice upon Tdepletion ( Figures 5 B, 5C, and S4 B). Increased granzyme B expression was not observed in NK cells in lungs, blood, and tumor upon Tdepletion despite higher baseline expression compared with Ax TDLNs ( Figures S4 C–S4E), indicating that tumor-activated Tinterfere with granzyme B expression of NK cells specifically in the Ax LN niche. Next, we analyzed the surface expression of CD107a on NK cells as a readout for their degranulation, which is an important mechanism for NK cell cytotoxicity (). This showed that NK cells in Ax TDLNs, but not lungs or Ax NDLNs, from T-depleted mice cells have increased surface expression of CD107a compared with control treatment ( Figures 5 D and S4 F), showing that Ax TDLN NK cells increase the release of intracellular granules upon Tdepletion in vivo. In in vitro stimulated NK cells, CD107a expression was not significantly affected by Tdepletion ( Figure S4 G), suggesting that the impact of Ton NK cell degranulation is not affecting their intrinsic capacity to degranulate under highly stimulatory conditions, but is rather a result of T/NK cell interactions in vivo. In contrast, we did not find a significant effect of Tdepletion on T cells in terms of granzyme B, CD107a, IFNγ, tumor necrosis factor alpha (TNFα) expression, or IFNγ by NK cells in Ax TDLNs and lungs ( Figures S4 H–S4L).

To gain more insight into how Tpromote metastasis formation in Ax LNs, we first explored the potential role of CD8T cells in controlling LN metastasis formation upon Tdepletion, as we found that Tsuppress IFNγ expression by CD8T cells in the primary tumor microenvironment ( Figure 4 D). To do so, we co-depleted CD8T cells ( Figure S4 A) and Tin the KEP metastasis model, but did not find a difference in LN metastasis incidence between anti-CD25 and anti-CD25/CD8 treatment ( Figure 5 A ), suggesting that the reduced LN metastasis incidence upon depletion of Tis not linked to intratumoral activation of CD8T cells ( Figure 4 D). Since we observed systemic activation and rewiring of highly immunosuppressive Tin response to tumorigenesis, we next hypothesized that Tmay differentially facilitate metastasis formation through tissue-specific interactions in the local metastatic niche, independent of their activity in the primary tumor.

(I) Venn diagram depicting distribution of upregulated genes (q < 0.1) between Ax TDLN and lung NK cells DT versus PBS treatment. Data in (C) and (D) show mean ± SEM p values determined by Mann-Whitney test (C and D) and Fisher's exact test (A).

(F and G) MA plot of differentially regulated transcripts for Ax TDLN NK cells (F), and lung NK cells (G) for DT versus PBS treatment. Significantly different transcripts are labeled in red (up), and blue (down).

(D) CD107a expression of NKp46 + DX5 + NK cells from Ax TDLNs and lungs of mice bearing transplanted KEP tumors (100 mm 2 ) receiving weekly neo-adjuvant treatment of 200 μg of anti-CD25 or mIgG2a (n = 6/group).

(C) Relative granzyme B expression by NK cells (CD3 − DX5 + NKp46 + ) in Ax TDLNs and lungs of mice bearing 100-mm 2 KEP tumors, treated with neo-adjuvant mIgG2a or anti-CD25, following a 3-h ex vivo stimulation (n = 6 mice/group). Data are normalized to % GzmB + of NK cells of control mIgG2a-treated mice.

(A) Percentage and number of mice with detectable micro-/macroscopic metastases in Ax TDLNs, in mice treated with neo-adjuvant indicated treatments (n = 11–18 mice/group).

After mastectomy, mice were monitored for the development of overt metastases. While neo-adjuvant Tdepletion did not improve metastasis-related survival or reduce the number of lung metastases ( Figures 4 F and S3 I), micro- and macroscopic analysis of Ax TDLNs ( Figure S3 J) revealed that anti-CD25-M2a-treated mice developed significantly fewer LN metastases compared with controls ( Figure 4 G). The incidence of LN metastasis of control mice was 93% (14 out of 15), which was reduced to 56% (9 out of 16) upon anti-CD25-M2a treatment. No difference was observed in the size of LN metastases that did develop in both groups ( Figure S3 K). The observation that Tdepletion reduces the incidence of LN metastasis by ∼50%, but does not affect lung metastasis, was consistent across four independent experimental KEP tumor donors, even though LN metastasis incidence of control groups varied between 41.67% and 93.3% in a donor-dependent fashion ( Figures 4 H and 4I). These findings indicate that Tpromote metastasis formation, leading to increased incidence of LN metastasis, but also reveals that the impact of Ton metastasis formation is dependent on the tissue context since lung metastases remain unaffected.

To assess the functional significance of Tduring early metastasis formation, we treated mice in the neo-adjuvant setting with a recently developed Fc-modified antibody, targeting the IL2Rα receptor, CD25 (anti-CD25-M2a), which has been described to efficiently and specifically deplete Tin tumors and peripheral tissue (). Indeed, anti-CD25-M2a treatment efficiently depleted FOXP3CD4T cells from tumors, spleen, LNs, lungs, and circulation in mice bearing transplanted KEP tumors ( Figures 4 B and 4C, S3 F, and S3G). Depletion of Twas observed for up to 10 days after start of treatment in blood. Although anti-CD25-M2a treatment resulted in increased interferon gamma (IFNγ) expression in both intratumoral CD4and CD8T cells ( Figure 4 D), consistent with the concept that tumor-induced Tare immunosuppressive, we did not observe an effect on primary tumor growth ( Figure 4 E). Similarly, depletion of Tin mammary tumor-bearing transgenic KEP mice did not affect primary tumor growth or survival ( Figure S3 H).

As we observed systemic and organ-specific mammary tumor-induced alterations of T, we set out to explore the impact of Ton multi-organ metastatic disease utilizing the KEP-based mastectomy model of spontaneous breast cancer metastasis ( Figure 4 A ) (). In this model, after orthotopic transplantation of a KEP-derived tumor fragment followed by surgical removal of the outgrown tumor, mice develop overt multi-organ metastatic disease, mainly in Ax TDLNs and lungs. Like primary tumor formation, metastatic disease is also accompanied by the accumulation of T, with elevated expression of ICOS, CTLA4, and ST2, compared with non-transplanted naive controls ( Figures S3 A–S3E).

(H and I) Ax TDLN metastasis incidence (H) and number of lung metastases (I) of each independent experimental donor is shown, in mice receiving weekly neo-adjuvant treatment of 200 μg of mIgG2a or anti-CD25. Symbol indicates an experimental group (mIgG2a/a-CD25), each line connects an independent experimental (donor #1 used in Figures 4 G and 4I, n = 15–16 mice/group; donor #2 used in Figure 5 A, n = 15–16; donor 3 and 4 used in Figure 6 A, n = 30–31 mice/group). Data in (B)–(D) and (F) show mean ± SEM p values determined by unpaired Student's t test (B, C, F), Mann-Whitney test (D), area under curve (AUC) calculation (E), Fisher's exact test (G), and paired Student's t test (H and I).

(G) Percentage and number of mice with detectable micro-/macroscopic metastases in Ax TDLNs, in mice treated with neo-adjuvant mIgG2a and anti-CD25 (n = 15–16 mice/group).

(D) Frequency of IFNγ + cells of CD4 + and CD8 + T cells, in 100-mm 2 mastectomized KEP tumors of mice treated with neo-adjuvant mIgG2a and anti-CD25 as determined by flow cytometry (n = 6–7 mice/group) following a 3-h ex vivo stimulation.

(B) Frequency of FOXP3 + cells of CD4 + T cells in mice bearing transplanted KEP tumors, treated with mIgG2a or anti-CD25 at indicated timepoints after start of treatment (n = 3–6 mice/group).

Taken together, these data demonstrate that the mammary tumor-induced changes in T regs are strongly influenced by their tissue context, raising the question of whether these site-specific differences may have functional consequences for the progression of breast cancer.

In addition to transcriptional commonalities observed in KEP Tin distant organs, we identified a large number of tumor-induced genes in KEP Tthat were not shared across multiple tissues, but rather were dependent on the tissue context ( Figure 2 G), indicating that the local environment shapes the response of Tto mammary tumorigenesis. Therefore, we continued our characterization of Tin distant organs of tumor-bearing KEP mice by exploring the impact of the tissue context. To do so, we omitted tumors and mammary glands from the dataset and re-analyzed the Ttranscriptome. PCA analysis revealed that Tderived from the same tissues cluster together, indicating that tissue residence is a more dominant factor for the transcriptional state of Tthan the presence or absence of a primary mammary tumor ( Figure 3 A ). To elaborate the relationship between Tin different tissues, we performed correlation analysis, and found Tfrom LNs, spleen, and blood to be relatively closely correlated, whereas lung Twere very distinct ( Figure 3 B). Interestingly, visualization of differentially regulated genes of matched tissues in a force-directed graph (KEP versus WT T, q < 0.05) () revealed complex relationships between clusters of genes dependent on the tissue context ( Figure 3 C). Among these, roughly 30% of differentially regulated genes in KEP Tversus WT Twere found to be tissue specific (74 out of 183 genes in LN, 379 out of 1,100 genes in lung, 417 out of 1,129 genes in blood, 107 out of 462 genes in spleen), indicating that KEP Tin distant organs acquire a unique tissue-specific transcriptional profile ( Figure 3 C, Table S1 ). We next performed IPA to interrogate which molecular pathways are associated with the differentially expressed genes between KEP and WT Tin distant organs. Notably, we identified several pathways related to T cell effector states (T-helper 1 [Th1] pathway, T-helper 2 [Th2] pathway, T helper cell differentiation, Th1 and Th2 activation pathway) to be shared among KEP Tin multiple distant organs ( Figure 3 D). In addition to shared pathways, we also found several pathways that were only observed for specific tissues, such as integrin signaling in blood Tand apoptosis signaling in lung T, highlighting the differential impact of mammary tumorigenesis on Tin distant organs.

(D) IPA on differentially expressed genes (q < 0.05) comparing T regs isolated from indicated tissues of KEP mice bearing end-stage tumors versus WT controls. Top 10 significant pathways are shown for each tissue.

(C) Force-directed graph depicting differentially expressed genes between KEP Tand WT T(q < 0.05). Genes identified by comparing gene expression of Tisolated from distant organs and blood of tumor-bearing KEP mice versus WT controls for each tissue, depicted by DiVenn ().

Taken together, these data demonstrate that mammary tumorigenesis induces systemic transcriptional rewiring of T regs , sharing a core set of genes associated with T reg function and activation.

We additionally identified Il1rl1, a gene encoding the IL-33 receptor ST2, to be systemically increased in KEP Tcompared with WT T, which was confirmed by FACS analysis ( Figures S2 B and S2C). Interleukin (IL)-33/ST2 signaling on Thas recently been described to induce a pro-tumorigenic phenotype in intratumoral T) and has also been shown to drive expansion of Tin vitro and in vivo (). In KEP mice, IL-33 was found to be significantly increased in TDLNs compared with WT LNs, which was not observed in blood, tumor, or lungs ( Figure S2 D). Nevertheless, short-term neutralization of IL-33 in tumor-bearing KEP mice utilizing two independent approaches, i.e., treatment of mice with anti-IL-33 or with an IL-33 antagonist (IL-33 Trap,) ( Figure S2 E), did not alter systemic Taccumulation, proliferation, or phenotype ( Figures S2 F–S2I), suggesting that, in mice with established mammary tumors, the presence of the ST2population is maintained independent of endogenous IL-33.

Next, we sought to explore how mammary tumorigenesis affects Tin distant organs by comparing gene expression profiles of KEP versus WT Tfrom matched tissues. This comparison identified differential gene regulation in Tin all organs tested, indicating that mammary tumors induce systemic transcriptional changes in T Figure 2 F). To further map these differentially regulated genes and their occurrence across different tissues, we analyzed their distribution in KEP versus WT Tacross tissues ( Figure 2 G). Doing so, we identified a set of 31 core genes to be significantly different (27 upregulated, three downregulated, one bidirectional, dependent on tissue) in KEP Tregardless of tissue residence, suggesting a certain level of convergent, tissue-independent transcriptional rewiring in response to mammary tumorigenesis ( Figure 2 H). Among those upregulated, we found genes encoding proteins important for T cell activation and the immunosuppressive features of T, such as Icos, Klrg1, Havcr2, Tigit, and Tnfrsf9. KEP Twere also found to have enhanced gene expression of Gzmb, which is known for its cytolytic function in NK and CD8T cells, but has been shown to contribute to immunosuppression when expressed by T). Combined, these data suggest that mammary tumorigenesis enhances systemic immunosuppression through transcriptional rewiring of Tin distant organs.

To delineate the impact of mammary tumor progression on Tin distant organs, RNA sequencing (RNA-seq) was performed on T(CD4CD25) isolated from blood, TDLNs, lungs, spleens, healthy mammary glands, and mammary tumors (225 mm) from tumor-bearing KEP mice and WT controls ( Figure 2 A ). Importantly, CD4CD25cells isolated from these tissues showed high and equal FOXP3 expression ( Figure S2 A). Principal component analysis (PCA) showed distinct clustering of T, based on their residence in either lymphoid tissue (spleen and LNs) and blood, or residence in peripheral tissue (lungs, tumor, mammary gland) ( Figure 2 B). Furthermore, Tresiding in distant organs cluster together independent of tumor status, whereas the gene expression profiles of tumor and mammary gland Tappear very distinct. Indeed, differential gene expression analysis comparing intratumoral KEP Tand mammary tissue-resident Trevealed 3,707 differentially expressed genes ( Figure 2 C). Ingenuity pathway analysis (IPA) showed the significantly changed pathways between Tfrom tumors and mammary glands to pertain to cell migration and extravasation ( Figure 2 D), which is underscored by some of the most differentially expressed genes, including Mmp10, Mmp13, and Ccr8 ( Table S1 ). We confirmed by gene set enrichment analysis (GSEA) that intratumoral KEP Tare significantly enriched for a clinically relevant cross-species and cross-tumor model tumor-infiltrating T(TIT) signature () ( Figure 2 E).

(F) Volcano plots showing differentially expressed genes (q < 0.05) from T regs isolated from indicated tissues of tumor-bearing KEP mice versus WT controls. Red indicates upregulated in KEP, blue indicates upregulated in WT.

We next determined the dynamics of Taccumulation and education by following Tfrequency and phenotype in aging KEP mice (from 2 to 8 months of age). Around 3 months of age, most KEP mice display microscopic neoplastic lesions in their mammary glands, which, over time, progress into palpable mammary tumors, with a median latency of 6–8 months (). Tfrequency in blood gradually increased during neoplastic progression in KEP mice, and was significantly increased in KEP mice of 7 months and older prior to the onset of palpable mammary tumors, compared with age-matched controls ( Figure 1 K). Further analysis of these Tshowed that the impact of mammary tumorigenesis on Tphenotype showed different kinetics per protein. Although the expression of CTLA4 increased prior to the development of palpable tumors, the expression of ICOS and CD103 was exclusively increased in tumor-bearing KEP mice ( Figures 1 L, S1 O, and S1P). Together, these data demonstrate that primary mammary tumorigenesis engages Tbeyond the tumor microenvironment, leading to their systemic expansion and activation.

Somatic inactivation of E-cadherin and p53 in mice leads to metastatic lobular mammary carcinoma through induction of anoikis resistance and angiogenesis.

Using high-dimensional flow cytometry, we observed that Tboth in- and outside of mammary tumors have increased expression of surface proteins associated with Tactivation and suppressor function including cytotoxic T lymphocyte-associated protein 4 (CTLA4), inducible T cell co-stimulator (ICOS), and CD103 in KEP tumor-bearing mice compared with WT controls, showing that these cells undergo a profound phenotypic change during mammary tumor progression ( Figures 1 F, S1 M, and S1N). To address whether the enhanced activation state of KEP Taffects their functionality throughout the tumor-bearing host, we used fluorescence-activated cell sorting (FACS) to sort Tfrom TDLNs, spleen, and tumors from KEP mice and WT controls to assess their suppressive activity on the proliferation of CD4and CD8T cells in vitro. Regardless of the tissue of origin, Tfrom tumor-bearing mice were significantly more potent in suppressing T cell proliferation compared with Tisolated from WT mice ( Figures 1 G–1J), indicating that tumor-educated Thave enhanced immunosuppressive potential, both intratumorally as well as in TDLNs and spleen.

To investigate whether this systemic increase of Tis consistently observed across preclinical mouse models of breast cancer, we analyzed Tfrequency in five different transgenic mouse models that represent different subsets of human breast cancers ( Figure 1 E). Indeed, we found Tto be significantly increased in the blood of tumor-bearing mice of all five models compared with WT controls, indicating systemic Texpansion is a prevalent feature of mammary tumorigenesis.

We then assessed whether Texpansion is explained by their increased proliferation or survival in tumor-bearing KEP mice. Ki67 expression on Tin tumor-bearing KEP mice was found to be uniquely increased in LNs, compared with WT controls ( Figure S1 J). Notably, no difference was observed between TDLNs and NDLNs, showing Tproliferation is systemically increased in LNs of tumor-bearing KEP mice ( Figure S1 K). Furthermore, KEP Tshowed increased viability when exposed to serum obtained from tumor-bearing KEP mice, as opposed to serum obtained from WT mice ( Figure S1 L). Combined, these data suggest that LNs may be an important site for Tproliferation in KEP mice, and that a soluble factor in KEP serum may contribute to increased Tsurvival.

To assess whether de novo mammary tumor formation exerts a systemic impact on T, we examined the abundance, phenotype, and activation status of Tin tumors, blood, and distant organs of the KEP mouse model, which spontaneously develops mammary tumors at 6–8 months of age resembling human invasive lobular carcinomas (ILCs) (). We observed that mammary KEP tumors are highly infiltrated by FOXP3CD4T cells, compared with healthy mammary glands of age-matched wild-type (WT) littermate controls ( Figures 1 A and 1B ). Interestingly, increased frequencies and absolute counts of Twere also observed in blood and in loco-regional or distant organs that are conducive to metastatic spread such as tumor-draining LNs (TDLNs; axillary and inguinal, dependent on the location of the primary mammary tumor), spleen, lungs, and non-draining LNs (NDLNs) of KEP mice bearing end-stage mammary tumors (225 mm) ( Figures 1 A, 1C, S1 A, and S1B). Notably, we did not find a relative increase in CD4FOXP3, or CD8T cells (with the exception of CD8T cells in TDLNs) in tumor-bearing KEP mice ( Figures 1 D, S1 C, and S1D). An increase in absolute cell counts was also observed for CD4FOXP3and CD8T cells in LNs and tumors ( Figures S1 E–S1G), due to expansion of these tissue compartments in KEP mice versus WT controls. However, comparing the ratio of FOXP3/CD8and FOXP3/FOXP3cells in different tissues of tumor-bearing KEP mice and WT controls ( Figures S1 H and S1I) confirmed that mammary tumorigenesis specifically and systemically expands Tamong the adaptive immune cell compartment.

(L) Frequencies of CTLA4 + cells of FOXP3 + CD4 + T cells in blood of tumor-free, tumor-bearing (225 mm 2 ) KEP mice and WT controls (n = 3–7 mice/group). Data in (B)–(F), (H)–(L) show mean ± SEM p values determined by unpaired Student's t test (B, D, H, I, J), one-way ANOVA with Dunnett's multiple comparison test (E), two-way ANOVA with Sidak's multiple comparison test (C and F), and Kruskal-Wallis test with Dunn's multiple comparison test (K and L). Asterisks indicate statistically significant differences compared with WT. ∗ p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001, ∗∗∗∗ p < 0.0001.

(H–J) Quantification of undivided responder cells (CD8 + and CD4 + T cells) based on CTV expression, upon co-culture with CD4 + CD25 + isolated from indicated tissues at various ratios (data pooled from three to four independent experiments, with two technical replicates per experiment).

(G) Representative histogram plots of Cell Trace Violet (CTV) expression in activated CD4/CD8 T cells alone (black) or upon co-culture with CD4 + CD25 + cells (gray and blue) obtained from indicated tissues at 1:2 T reg :responder ratio.

(F) Representative histograms depicting expression (left) and quantification (right) of CTLA4, ICOS, and CD103 gated on CD4 + FOXP3 + T cells, in indicated tissues of KEP mice (blue) bearing tumors (225 mm 2 ) versus WT littermates (black) by flow cytometry (n = 3–11 mice/group).

(E) Frequencies of FOXP3 + cells of CD4 + T cells in blood of mice bearing end-stage tumors of indicated transgenic mouse models for mammary tumorigenesis compared with age-matched WT mice (n = 8–22 mice/group).

(B and C) Frequencies of FOXP3 + cells of CD4 + T cells in indicated tissues of KEP mice bearing mammary tumors (225 mm 2 ) versus WT controls (n = 6–15 mice/group) as determined by flow cytometry.

Somatic inactivation of E-cadherin and p53 in mice leads to metastatic lobular mammary carcinoma through induction of anoikis resistance and angiogenesis.

Discussion

Understanding the nature of cancer-associated systemic immunosuppression and its impact on different (pre-)metastatic niches is essential to ultimately design effective therapeutic strategies that prevent or fight metastatic disease. Here, we show that mammary tumorigenesis has an extensive impact on T regs , both intratumorally and in distant organs. Tumor-educated T regs are highly activated and immunosuppressive, and display tissue-specific adaptation to tumor development. This has functional relevance for metastasis formation, as T regs selectively promote LN metastasis, but not lung metastasis, through inhibition of NK cells. These data highlight the importance of the tissue context for immune escape mechanisms, and reveal a causal role for T regs in the development of LN metastasis.

Huntington et al., 2020 Huntington N.D.

Cursons J.

Rautela J. The cancer–natural killer cell immunity cycle. López-Soto et al., 2017 López-Soto A.

Gonzalez S.

Smyth M.J.

Galluzzi L. Control of metastasis by NK cells. regs . The relevance of these findings for human breast cancer is supported by our observation that the T reg /NK cell ratio strongly shifts toward T regs in SLNs of BrCa patients compared with healthy LNs. In addition, an explorative study using metastatic LNs from melanoma patients showed that ex vivo depletion of T regs enhanced cytolytic activity of LN NK cells in vitro, suggesting T regs can also inhibit LN NK cells in melanoma ( Ghiringhelli et al., 2005 Ghiringhelli F.

Ménard C.

Terme M.

Flament C.

Taieb J.

Chaput N.

Puig P.E.

Novault S.

Escudier B.

Vivier E.

et al. CD4+CD25+ regulatory T cells inhibit natural killer cell functions in a transforming growth factor-β-dependent manner. reg accumulation ( Mamessier et al., 2011 Mamessier E.

Sylvain A.

Thibult M.L.

Houvenaeghel G.

Jacquemier J.

Castellano R.

Gonçalves A.

André P.

Romagné F.

Thibault G.

et al. Human breast cancer cells enhance self tolerance by promoting evasion from NK cell antitumor immunity. Kim et al., 2019 Kim R.

Kawai A.

Wakisaka M.

Funaoka Y.

Yasuda N.

Hidaka M.

Morita Y.

Ohtani S.

Ito M.

Arihiro K. A potential role for peripheral natural killer cell activity induced by preoperative chemotherapy in breast cancer patients. reg immunosuppression ( Wing et al., 2008 Wing K.

Onishi Y.

Prieto-Martin P.

Yamaguchi T.

Miyara M.

Fehervari Z.

Nomura T.

Sakaguchi S. CTLA-4 control over Foxp3+ regulatory T cell function. NK cells are a well-recognized key element of the anti-tumor response (), but the role and cellular crosstalk of NK cells in the context of LN metastasis has remained unclear. Here, we show that NK cells in Ax TDLNs have anti-metastatic potential, provided they are relieved from the immunosuppressive pressure by T. The relevance of these findings for human breast cancer is supported by our observation that the T/NK cell ratio strongly shifts toward Tin SLNs of BrCa patients compared with healthy LNs. In addition, an explorative study using metastatic LNs from melanoma patients showed that ex vivo depletion of Tenhanced cytolytic activity of LN NK cells in vitro, suggesting Tcan also inhibit LN NK cells in melanoma (). For breast cancer specifically, the expression of granzyme B within tumor-infiltrating NK cells was found to negatively correlate to Taccumulation (). Furthermore, another recent study identified that clearance of Ax LN metastasis in breast cancer patients treated with neo-adjuvant chemotherapy significantly correlated with increased cytotoxic potential of NK cells in peripheral blood as well as with decreased intratumoral CTLA4 gene expression (), which is well known to be important for Timmunosuppression ().

regs show tissue-dependent rewiring in response to mammary tumorigenesis, which may either be explained through tissue-specific upstream regulators or is reflective of the distinct inherent differences between tissue-resident T regs , in particular in lymphoid versus non-lymphoid organs. Tissue context not only drives T reg phenotype in tumor-bearing hosts but also dictates the interaction between T regs and one of their cellular targets, NK cells. Specifically, we found T reg depletion to differentially affect lung and LN NK cells at both the transcriptional and the protein level ( reg depletion, we found this to unleash NK cell-mediated anti-metastatic activity only in the LN niche, but not the lungs. We speculate that this may occur through the induction of an effector mechanism observed specifically in LN—but not lung—NK cells, such as increased expression of the cytotoxic molecules granzyme A and B, NK cell co-stimulatory receptors, or chemokine receptors ( reg depletion are intrinsic to LN NK cells or due to a unique feature of tumor-educated T regs in TDLNs remains to be elucidated. Alternatively, NK cells may be functionally repressed through other immunosuppressive mechanisms that are independent of T regs , and specific to the lung niche. For example, we observed that lung, but not LN, NK cells undergo a shift toward a more immature state in tumor-bearing mice ( regs and potentially affects their anti-metastatic potential. In line with this hypothesis, a recent study revealed that lung NK cells are suppressed by IL-33 activated innate lymphoid type 2 cells, which stunts their ability to control pulmonary metastasis of intravenously injected B16F10 cells ( Schuijs et al., 2020 Schuijs M.J.

Png S.

Richard A.C.

Tsyben A.

Hamm G.

Stockis J.

Garcia C.

Pinaud S.

Nicholls A.

Ros X.R.

et al. ILC2-driven innate immune checkpoint mechanism antagonizes NK cell antimetastatic function in the lung. regs in the lungs, highlighting the importance of local, tissue-specific mechanisms of immunosuppression and cancer immune surveillance during metastasis formation. Finally, NK cells may be functionally irrelevant for lung metastasis formation in this model independent of their activation status, through cancer cell-intrinsic differences between lymph and lung metastasizing cancer cells that affects their likelihood to be killed by NK cells ( López-Soto et al., 2017 López-Soto A.

Gonzalez S.

Smyth M.J.

Galluzzi L. Control of metastasis by NK cells. Reiter et al., 2020 Reiter J.G.

Hung W.T.

Lee I.H.

Nagpal S.

Giunta P.

Degner S.

Liu G.

Wassenaar E.C.E.

Jeck W.R.

Taylor M.S.

et al. Lymph node metastases develop through a wider evolutionary bottleneck than distant metastases. Our findings demonstrate that Tshow tissue-dependent rewiring in response to mammary tumorigenesis, which may either be explained through tissue-specific upstream regulators or is reflective of the distinct inherent differences between tissue-resident T, in particular in lymphoid versus non-lymphoid organs. Tissue context not only drives Tphenotype in tumor-bearing hosts but also dictates the interaction between Tand one of their cellular targets, NK cells. Specifically, we found Tdepletion to differentially affect lung and LN NK cells at both the transcriptional and the protein level ( Figure 5 ). Although transcriptomic analyses revealed that LN and lung NK cells acquire a more activated phenotype upon Tdepletion, we found this to unleash NK cell-mediated anti-metastatic activity only in the LN niche, but not the lungs. We speculate that this may occur through the induction of an effector mechanism observed specifically in LN—but not lung—NK cells, such as increased expression of the cytotoxic molecules granzyme A and B, NK cell co-stimulatory receptors, or chemokine receptors ( Figure 5 I). Whether these specific phenotypic alterations observed in LN NK cells upon Tdepletion are intrinsic to LN NK cells or due to a unique feature of tumor-educated Tin TDLNs remains to be elucidated. Alternatively, NK cells may be functionally repressed through other immunosuppressive mechanisms that are independent of T, and specific to the lung niche. For example, we observed that lung, but not LN, NK cells undergo a shift toward a more immature state in tumor-bearing mice ( Figures S5 D and S5E). This occurs independent of Tand potentially affects their anti-metastatic potential. In line with this hypothesis, a recent study revealed that lung NK cells are suppressed by IL-33 activated innate lymphoid type 2 cells, which stunts their ability to control pulmonary metastasis of intravenously injected B16F10 cells (). This shows that NK cells can be suppressed beyond the control of Tin the lungs, highlighting the importance of local, tissue-specific mechanisms of immunosuppression and cancer immune surveillance during metastasis formation. Finally, NK cells may be functionally irrelevant for lung metastasis formation in this model independent of their activation status, through cancer cell-intrinsic differences between lymph and lung metastasizing cancer cells that affects their likelihood to be killed by NK cells (), which we have not explored in this study.

+ T regs ( Hemmers et al., 2021 Hemmers S.

Schizas M.

Rudensky A.Y. T reg cell-intrinsic requirements for ST2 signaling in health and neuroinflammation. reg accumulation or proliferation ( reg expansion in mammary tumor-bearing mice is regulated independent of IL-33, and thus remains an avenue of future research. Our in vitro studies ( reg survival. An important cytokine involved in T reg proliferation and survival ( Sun et al., 2019b Sun Z.

Ren Z.

Yang K.

Liu Z.

Cao S.

Deng S.

Xu L.

Liang Y.

Guo J.

Bian Y.

et al. A next-generation tumor-targeting IL-2 preferentially promotes tumor-infiltrating CD8+ T-cell response and effective tumor control. reg expansion, as has been observed in tumor-bearing mice treated with recombinant IL-2 ( Sun et al., 2019b Sun Z.

Ren Z.

Yang K.

Liu Z.

Cao S.

Deng S.

Xu L.

Liang Y.

Guo J.

Bian Y.

et al. A next-generation tumor-targeting IL-2 preferentially promotes tumor-infiltrating CD8+ T-cell response and effective tumor control. Despite observations that exogenous IL-33 can induce acute peripheral accumulation of ST2), we find that blockade of endogenous IL-33 does not affect tumor-induced systemic Taccumulation or proliferation ( Figures S2 F–S2I), suggesting that Texpansion in mammary tumor-bearing mice is regulated independent of IL-33, and thus remains an avenue of future research. Our in vitro studies ( Figure S1 L) suggest that a soluble factor in KEP serum can promote Tsurvival. An important cytokine involved in Tproliferation and survival () that we did not study here is IL-2. Therefore, future studies may analyze whether IL-2 is differentially expressed or regulated in tumor-bearing hosts, and might contribute to Texpansion, as has been observed in tumor-bearing mice treated with recombinant IL-2 ().

In conclusion, these findings reveal a causal role for T regs in the formation of LN metastasis through local suppression of NK cells, and may form the basis for the design of neo-adjuvant therapeutic strategies aimed to reduce nodal metastasis in breast cancer patients.