Abstract Neutrophils, the most abundant innate immune cells, function as crucial regulators of the adaptive immune system in diverse pathological conditions, including metastatic cancer. However, it remains largely unknown whether their immunomodulatory functions are intrinsic or acquired within the pathological tissue environment. Here, using mouse models of metastatic breast cancer in the lungs, we show that, while neutrophils isolated from bone marrow (BM) or blood are minimally immunosuppressive, lung-infiltrating neutrophils are robustly suppressive of both T cells and natural killer (NK) cells. We found that this tissue-specific immunosuppressive capacity of neutrophils exists in the steady-state, and is reinforced by tumor-associated inflammation. Acquisition of potent immunosuppression activity by lung-infiltrating neutrophils was endowed by the lung-resident stroma, specifically CD140a^+ mesenchymal cells (MCs) and largely via prostaglandin-endoperoxide synthase 2 (PTGS2), the rate-limiting enzyme for prostaglandin E2 (PGE2) biosynthesis. MC-specific deletion of Ptgs2 or pharmacological inhibition of PGE2 receptors substantially reversed lung neutrophil-mediated immunosuppression and mitigated lung metastasis of breast cancer in vivo. Importantly, these lung stroma-targeting strategies substantially improved the therapeutic efficacy of adoptive T cell-based immunotherapy in treating metastatic disease in mice. Collectively, our results reveal that the immunoregulatory effects of neutrophils are induced by tissue-resident stroma, and that targeting tissue-specific stromal factors represents an effective approach to boost tissue-resident immunity against metastatic disease. One Sentence Summary: Lung-resident mesenchymal cells reprogram neutrophils to be immunosuppressive cells that facilitate lung metastasis of breast cancer. INTRODUCTION Most solid cancer-related deaths are attributed to metastatic disease developing in distant vital organs ([36]1). Metastasis is a complex process involving multiple sequential steps, including tumor cell invasion from the primary tumor, intravasation, survival in the circulation, extravasation, and organ colonization. It is widely accepted that metastatic colonization is the most inefficient step, as the vast majority of disseminated tumor cells (DTCs) undergo apoptosis shortly after they arrive at the secondary sites ([37]1, [38]2). Very few DTCs are capable of bypassing local immune surveillance to achieve successful colonization ([39]3–[40]5). Accumulating evidence from studies of metastatic colonization has revealed a crucial role for the organ microenvironment in fostering the colonization and outgrowth of DTCs during formation of metastatic lesions ([41]6–[42]8). Thus, a deeper characterization of the organ microenvironment is imperative for defining the precise mechanisms of metastasis and for enabling the design of organ-specific therapeutic interventions to treat metastatic disease. The lungs are one of the most common sites of metastasis in many solid cancer types. Within the lung microenvironment, bone marrow (BM)-derived neutrophils, the most abundant innate immune cells, are an indispensable component in modulation of lung metastases ([43]9–[44]15). The metastasis-promoting effects of neutrophils are mainly achieved through their suppression of lung-resident anti-tumor immunity ([45]13, [46]15–[47]18), which creates a hospitable environment for invading DTCs. In addition to their immunosuppressive capacity, neutrophils facilitate metastasis by awaking dormant cancer cells, accelerating DTC extravasation, enhancing circulating tumor cell proliferation, and promoting angiogenesis ([48]9, [49]10, [50]19–[51]23). Together with other lung cellular components, neutrophils elicit the formation of the pre-metastatic and metastatic niches in the lung. In contrast to extensive studies on neutrophil-mediated immunosuppression at primary tumor sites, the mechanisms underlying neutrophil modulation of anti-tumor immunity within metastatic organs are less well characterized ([52]17, [53]24, [54]25). Limited evidence shows that, at the pre-metastatic stage, lung-infiltrating neutrophils are capable of suppressing CD8^+ T cells and natural killer (NK) cells via inducible nitric oxide synthase (iNOS) or reactive oxygen species (ROS) ([55]13, [56]15). However, a fundamental question remains unanswered – whether the immunosuppressive capacity of neutrophils exists intrinsically when they originate in the hematopoietic organs and circulate in blood, or whether it is acquired after they infiltrate into the lung microenvironment. In this study, we therefore compared the immunosuppressive profiles of neutrophils across different tissues under both steady-state and tumor-bearing conditions, and determined the cellular and molecular mechanisms driving the tissue-specific characteristics of neutrophils. RESULTS The immunosuppressive capacity of neutrophils is tissue-dependent under both steady-state and tumor-bearing conditions To determine whether the immunosuppression exerted by neutrophils is intrinsic or elicited by local environments of metastatic organs, we compared neutrophils residing in the hematopoietic site (i.e., BM), circulating in the peripheral blood (PB), and those infiltrating the metastatic organ (lung) at the pre-metastatic stage ([57]fig. S1A) of orthotopic 4T1 tumor-bearing mice – a breast tumor model characterized by induction of potent neutrophilic inflammation ([58]26) ([59]fig. S1, B–[60]D). Functionally, neutrophils (CD45^+CD11b^+Ly6C^lowLy6G^+; [61]fig. S1B) isolated from pre-metastatic lungs had a more robust ability to suppress CD4^+ and CD8^+ T cell proliferation ([62]Fig. 1, A and [63]B, and [64]fig. S1E), and NK cell cytotoxicity ([65]fig. S1, E and [66]F), than their BM and PB counterparts. Through RNA-sequencing (RNA-seq), lung neutrophils were found to express remarkably higher levels of immunosuppression-associated genes, including Ptgs2, Il1b, Il10, Arg1, Arg2, Nos2, Cd274, Trem1, Fas, Nfe2l2 and Pdcd1lg2 ([67]27–[68]30), than BM or PB neutrophils isolated from 4T1 tumor-bearing mice ([69]Fig. 1, C and [70]D). In addition, certain signature genes expressed by myeloid-derived suppressor cells (MDSCs), including Vegfa, Cd14, Clec4d, Clec4e and C5ar1 ([71]20, [72]31, [73]32), were also highly expressed by lung neutrophils ([74]Fig. 1C). Similar to the 4T1 model, a lung neutrophil-enriched immunosuppressive gene expression pattern was observed in the orthotopic AT3 model ([75]fig. S2A). At the protein level, we consistently detected notably higher expression of immunosuppression-associated factors in lung neutrophils than in BM or PB neutrophils isolated from 4T1 tumor-bearing mice ([76]Fig. 1E). Intriguingly, this tissue-specific expression of immunosuppression-associated genes by neutrophils was not only detected under tumor-bearing conditions, but was also seen in tumor-free naïve mice ([77]fig. S2B). Fig. 1. The immunosuppressive capacity of neutrophils is tissue-dependent under both steady-state and tumor-bearing conditions. Fig. 1. [78]Open in a new tab (A-B) As depicted in the schematic (A), neutrophils (Neu) (CD45^+CD11b^+Ly6C^lowLy6G^+) were isolated from BM, PB and lung of 4T1 tumor-bearing mice, and then co-cultured with splenic CD3^+ T cells at a ratio of 1:1 in the presence of plate-coated anti-CD3 and soluble anti-CD28. T cell proliferation was quantified by flow cytometry (B) (n=4). (C-D) RNA-seq analysis of BM, PB and lung neutrophils from 4T1 tumor-bearing mice (n=3). (C) Heatmap showing the expression of the indicated immunosuppression-associated genes from the RNA-seq data. (D) Volcano plots showing fold change and P-value for the comparison of gene expression in lung neutrophils versus PB neutrophils. The genes of interest are indicated in volcano plots. (E) Representative immunofluorescence images of BM, PB and lung neutrophils from 4T1 tumor-bearing mice, stained for interleukin 1 beta (IL-1β), interleukin 10 (IL-10), arginase 2 (ARG2), and nitric oxide synthase 2 (NOS2). Scale bars, 10 μm. (F) Fluorescent dye-labeled, BM-derived healthy neutrophils were intravenously (IV) injected into tumor-free recipient mice. 16 hours later, the indicated tissues were harvested, and the implanted neutrophils were sorted. Relative mRNA expression of indicated immunosuppression-associated genes was measured by quantitative PCR (n=4). (G) Time-dependent analyses of implanted exogenous neutrophils in PB and lung at the indicated times post-transfer (n=4). n represents the number of biological replicates. The results in B and E-G are representative of three independent experiments and shown as mean ± SEM. Statistical significance was determined by one-way ANOVA with Tukey’s multiple comparisons test (B, F) or two-way ANOVA with Sidak’s multiple comparisons test (G). *p< 0.05; **p< 0.01; ***p< 0.001; ****p< 0.0001; NS, not significant. As the spleen has been reported to act as a reservoir to supply immunosuppressive and tumor-promoting neutrophils in various types of cancers ([79]31, [80]33, [81]34), we then compared lung- and spleen-derived neutrophils directly. Under both steady-state and tumor-bearing conditions, lung neutrophils expressed higher levels of immunosuppression-associated genes and accordingly exerted more potent inhibitory effects on T cell proliferation and NK cell cytotoxicity, compared to neutrophils isolated from spleens under the same tissue dissociation conditions ([82]fig. S2, C–[83]E). This was further supported by analysis of a published dataset ([84]GSE142432) ([85]35), in which we found that lung neutrophils express a higher level of immunosuppression-associated genes than their spleen counterparts isolated from either germ-free or specific-pathogen free naïve mice ([86]fig. S2F). Collectively, lung neutrophils display a distinctly robust expression of immunoregulatory factors in both naïve and tumor-bearing mice. To test whether the lung environment serves to reprogram infiltrated neutrophils to acquire an immunosuppressive phenotype, we intravenously (IV) injected fluorescent dye-labeled, BM-derived healthy neutrophils into naïve recipient mice. Upon analysis of transplanted neutrophils engrafted in different tissues, we found that exogenous neutrophils infiltrating the lung, but not those that infiltrated other tissues (BM, PB, spleen, and liver), substantially increased their expression of immunosuppression-associated genes ([87]Fig. 1F and [88]fig. S3A), indicating a lung-specific capacity to reprogram infiltrated neutrophils. Time-dependent analyses of exogenous neutrophils in the PB and lungs of naïve recipient mice revealed a rapid acquisition of the immunosuppressive phenotype by lung-infiltrating neutrophils as early as 4 hours post-transfer ([89]Fig. 1G and [90]fig. S3B). When 4T1 tumor-bearing mice were employed as recipients, this acquired immunosuppressive characteristic of implanted neutrophils was further reinforced, compared to naïve recipients ([91]fig. S3C). Thus, the lung environment has an intrinsic capacity to quickly reprogram infiltrated neutrophils to become immunosuppressive. Previous studies have shown that the lung accommodates neutrophils with prolonged transit within the pulmonary vasculature under steady-state ([92]36), and that such a marginated pool of neutrophils can be mobilized from the lung to the circulation by plerixafor, an antagonist for C-X-C chemokine receptor type 4 (CXCR4) ([93]37). Indeed, we found the number of circulating neutrophils was rapidly increased within 2 hours of plerixafor administration in naïve mice, which was gradually reduced afterwards ([94]fig. S3, D and [95]E). Accordingly, the immunosuppression-associated gene expression in circulating neutrophils underwent a remarkable elevation after 1 hour upon plerixafor treatment, and then decreased thereafter, with most of the tested genes returning to original or near-original expression levels by 4 hours ([96]fig. S3F). These results suggested that the lung “immune suppressive” neutrophils could be mobilized into the circulation, which may return to the lung as the effect of plerixafor waned. The immunosuppressive property of lung neutrophils is not attributed to a particular subpopulation To characterize whether the lung-associated immunosuppressive property might pertain to particular neutrophil subpopulations, we performed single-cell RNA-sequencing (scRNA-seq) on BM, PB, and lung neutrophils isolated from both naïve (BALBc/J) and 4T1 tumor-bearing mice ([97]fig. S4A). Unbiased comparative analyses ([98]38) identified 10 clusters present in both naïve and tumor-bearing states ([99]Fig. 2A, left). In naïve mice, BM, PB, and lung neutrophils all displayed a high extent of tissue-specific gene expression ([100]fig. S4B) and tissue-specific clustering patterns ([101]Fig. 2A). Under tumor-bearing conditions, BM and PB neutrophils showed a tendency to be clustered together, while the cluster distribution of lung neutrophils remained largely unaltered vs. naïve state with no significant overlap with BM or PB neutrophil clusters ([102]Fig. 2A), indicating a tissue-restricted transcriptional portrait of lung neutrophils. Fig. 2. The immunosuppressive property of lung neutrophils is not attributed to a particular subpopulation. Fig. 2. [103]Open in a new tab (A) t-SNE plots showing the distribution of BM, PB and lung neutrophils from naïve and 4T1 tumor-bearing mice (n=1, no sample pooling). Each point represents a single cell colored according to cluster designation. Lung neutrophils are indicated, and plots at right highlight cells from each individual tissue. (B) Pie charts showing the percentage of three lung neutrophil subsets (clusters 2, 6 and 7) in naïve and 4T1 tumor-bearing mice. (C) The expression of indicated genes on the t-SNE plots from the data in (A). (D) Representative immunofluorescence images of lung sections from naïve and 4T1 tumor-bearing mice, stained for Ly6G (red). Scale bars, 50 μm. (E) The absolute number of lung neutrophils was measured in naïve and 4T1 tumor-bearing mice (n=4). (F) Relative mRNA expression of the indicated genes was measured in lung neutrophils from naïve and 4T1 tumor-bearing mice (n=4). (G-H) Lung neutrophils were isolated from naïve and 4T1 tumor-bearing mice, and then their immunosuppressive effects on T cells (G) or NK cells (H) were measured (n=4). Neutrophil: T or NK cell = 1:1. n represents the number of biological replicates. The results in D-H are representative of three independent experiments and shown as mean ± SEM. Statistical significance was determined by unpaired two-tailed Student’s t-test (E-F) or one-way ANOVA with Tukey’s multiple comparisons test (G-H). *p< 0.05; **p< 0.01; ***p< 0.001; ****p< 0.0001. Lung neutrophils were mainly comprised of three distinct subpopulations: cluster 2 (C2), cluster 6 (C6), and cluster 7 (C7) ([104]Fig. 2A and [105]fig. S4C). In the steady-state, C2 and C7 accounted for ~95% of all lung neutrophils, while C6 represented only a minor portion (~5%) ([106]Fig. 2B, top). With primary breast tumor progression, there was a striking >10-fold expansion of the C6 subpopulation in the pre-metastatic lungs, and an accompanying ~2-fold reduction for both C2 and C7 subpopulations ([107]Fig. 2B). Since 4T1 tumor cells produce a substantial amount of granulocyte colony-stimulating factor (G-CSF) ([108]26), which is known to induce egress of immature neutrophils from BM to circulation and tissues ([109]39, [110]40), we speculated that the expanded C6 subpopulation in the 4T1 model might be contributed by immature neutrophils. Employing CD101 as the surface marker to distinguish immature (CD101^−) and mature (CD101^+) neutrophils ([111]39, [112]40) ([113]fig. S4, D–[114]F), we found that CD101^− immature neutrophils were markedly increased in PB and lungs, but decreased in BM of the 4T1 tumor-bearing mice ([115]fig. S4, G–[116]H). Of note, lung CD101^− immature neutrophils expressed higher levels of the signature genes of the C6 cluster ([117]fig. S4C), including Prok2, Ifitm6, Lcn2, Aldh2, Stfa2, Ltf and Stfa3, compared to CD101^+ mature lung neutrophils ([118]fig. S4I). Meanwhile, the C6 cluster had a more pronounced expression of immature neutrophil signature genes than the other two clusters ([119]fig. S4J). Thus, the C6 subpopulation is likely enriched for immature-like neutrophils, and its expansion under tumor-bearing conditions could be attributed to G-CSF-mobilized immature neutrophils from BM to the lung. Next, we attempted to map the immunosuppression-associated genes into the above neutrophil clusters to determine whether the characteristic immunosuppressive phenotype of lung neutrophils belongs to certain subpopulation(s), particularly the tumor-associated immature-like C6 cluster. At the steady-state, the prominently expressed immunosuppression-associated genes in lung neutrophils identified through bulk RNA-seq, such as Ptgs2, Il1b, Arg2, Trem1, Nfe2l2, and Cd14, were indeed more abundantly expressed in the lung neutrophil clusters than clusters of BM or PB neutrophils ([120]Fig. 2C, top). Probing the three naive lung neutrophil clusters, we found the above signature genes were expressed in each cluster ([121]Fig. 2C, top); however, quantitative analysis showed that the C6 cluster had a relatively low expression of these genes among the three clusters ([122]fig. S4K, top). Under the 4T1 tumor-bearing condition, despite a drastic change in the frequency of the three clusters, the expression pattern of immunosuppression-associated genes among clusters was not significantly altered, with the immature-like neutrophil-enriched C6 having the lowest relative expression ([123]Fig. 2C, bottom and [124]fig. S4K, bottom). Consistent with the scRNA-seq data, sorted lung CD101^− immature neutrophils from the 4T1 tumor-bearing mice indeed expressed lower levels of immunosuppression-associated genes than the CD101^+ mature lung neutrophils ([125]fig. S4L). Thus, the immunosuppressive phenotype of lung neutrophils is likely not restricted to particular subpopulations under either naïve or tumor-bearing conditions, despite a quantitative difference among the three clusters. In subsequent experiments, accordingly, we studied lung neutrophils as a whole. We next sought to quantitatively define how the immunosuppressive property of lung neutrophils is modulated by tumor-bearing conditions. In the 4T1 model, along with a striking increase in absolute numbers of lung-infiltrating neutrophils at the pre-metastatic stage ([126]Fig. 2, D and [127]E), their expression of immunosuppression-associated genes was found to be remarkably elevated, as measured by quantitative PCR ([128]Fig. 2F). In line with these transcriptional changes, neutrophils isolated from the pre-metastatic lungs showed a more robust ability than naïve lung neutrophils to suppress both CD4^+ and CD8^+ T cell proliferation ([129]Fig. 2G), as well as the cytotoxicity of NK cells ([130]Fig. 2H). Therefore, lung neutrophil-specific immunosuppressive traits, which exist at the steady-state, can be reinforced by tumor-bearing conditions. Host neutrophilic inflammation strengthens the immunosuppressive capacity of lung neutrophils The 4T1 model used above has been reported to be associated with induction of profound host neutrophilic inflammation ([131]26). We therefore sought to distinguish the effect of tumor-bearing vs. host neutrophilic inflammation on reinforcement of lung neutrophil-specific immunosuppressive function. To this end, we employed the AT3 tumor cell line that originated from a MMTV-PyMT tumor (C57BL/6J background) and that induces relatively mild host inflammation ([132]41). In parallel, we capitalized on our newly constructed AT3-gcsf cell line that overexpresses G-CSF and stimulates robust host neutrophilic inflammation ([133]13) ([134]fig. S5A). We first profiled BM, PB, and lung neutrophils derived from naïve (C57BL/6J), AT3-, and AT3-gcsf orthotopic tumor-bearing mice using scRNA-seq ([135]fig. S5B). As expected, lung neutrophils displayed a tissue-specific clustering distribution across the three different host conditions ([136]Fig. 3A and [137]fig. S5C). Neither tumor-bearing nor host inflammation status significantly altered the tissue-specific immunosuppressive gene expression pattern of neutrophils ([138]Fig. 3B), or the expression profile of these genes within lung neutrophil clusters ([139]fig. S5D). Consistent with the 4T1 model, the C6 cluster, which we again identified to be enriched with immature-like neutrophils ([140]fig. S5, E–[141]G), was increased under AT3 tumor-bearing condition and further notably expanded by AT3-gcsf tumor-bearing condition ([142]Fig. 3C). In line with the scRNA-seq data, flow cytometric analysis showed a similar shift of lung CD101^− immature neutrophils across the three different host conditions ([143]fig. S5, H–[144]I), suggesting a G-CSF-triggered immature neutrophil emigration from BM to the lungs. Therefore, G-CSF-induced host neutrophilic inflammation did not influence the tissue-specific immunosuppressive gene expression profile of neutrophils, although it did cause lung neutrophil subpopulations to shift in frequency. Fig. 3. Host inflammation strengthens the immunosuppressive capacity of lung neutrophils. Fig. 3. [145]Open in a new tab (A) t-SNE plots showing the distribution of PB, BM and lung neutrophils from naïve, AT3 tumor-bearing and AT3-gcsf tumor-bearing mice (n=1, no sample pooling). Lung neutrophils are indicated. (B) Dot plots showing the expression of the indicated genes in PB, BM and lung neutrophils across different host conditions. (C) Pie charts showing the percentage of three lung neutrophil subsets (clusters 1, 3 and 6) in naïve, AT3 tumor-bearing and AT3-gcsf tumor-bearing mice. (D) Relative mRNA expression of the indicated genes was measured in lung neutrophils from naïve, AT3 tumor-bearing and AT3-gcsf tumor-bearing mice (n=4). (E-F) Lung neutrophils were isolated from naïve, AT3 tumor-bearing and AT3-gcsf tumor-bearing mice, and then their immunosuppressive effects on T cells (E) or NK cells (F) were measured (n=4). Neutrophil: T or NK cell = 1:1. (G) Representative immunofluorescence images of lung sections from AT3 tumor-bearing and AT3-gcsf tumor-bearing mice, stained for Ly6G (red) and CD3 (white) (left). Scale bars, 50 μm. The distance of T cell to nearest neutrophil was measured (right). (H) As depicted in the schematic ([146]Fig. S5C), lung metastatic colonization was detected by ex vivo bioluminescence imaging (BLI) and primary tumor weight was measured (n=10). n represents the number of biological replicates. The results in D-H are representative of three independent experiments and shown as mean ± SEM. Statistical significance was determined by one-way ANOVA with Tukey’s multiple comparisons test (D-F), unpaired two-tailed Student’s t-test (G) or unpaired two-tailed Mann-Whitney test (H). *p< 0.05; **p< 0.01; ***p< 0.001; ****p< 0.0001; NS, not significant. We next determined how host neutrophilic inflammation quantitatively affects the immunosuppressive characteristic of lung neutrophils. At the pre-metastatic stage, while the AT3 tumor-bearing condition mildly increased lung neutrophil expressions of the immunosuppression-associated genes, these genes underwent a further significant upregulation in the AT3-gcsf model ([147]Fig. 3D). Such a changing pattern of lung neutrophils at transcriptional level was reflected in their functional changes in suppressing T cells ([148]Fig. 3E) and NK cells ([149]Fig. 3F), indicating that host neutrophilic inflammation condition heightens the immunosuppressive capacity of lung neutrophils and may better elicit formation of the lung pre-metastatic niche. Supporting this conjecture, a lung tissue-specific infiltration of neutrophils ([150]fig. S6A) and a close in situ interaction between lung neutrophils and T cells ([151]Fig. 3G) were detected under the host neutrophilic inflammation condition (AT3-gcsf model). Moreover, in pre-metastatic lungs, T cells were physically closer to neutrophils than to CD11c^+ myeloid cells, which include macrophages and dendritic cells ([152]42), in AT3-gcsf tumor-bearing mice ([153]fig. S6B), suggesting that neutrophils may play a more dominant role than other myeloid cells in mediating immunosuppression under this robust neutrophilic inflammation. Using a modified experimental lung metastasis model ([154]fig. S6C), in which luciferase-labeled tumor cells were IV injected into orthotopic tumor-bearing mice, hence the role of pre-metastatic niche can be evaluated ([155]14, [156]43), we found that the strong host inflammation condition (primary AT3-gcsf tumors) was indeed superior to the weak one (primary AT3 tumors) in accommodating colonization of AT3-Luc cells ([157]Fig. 3H, left). In contrast, primary tumor growth did not significantly differ under these two host conditions ([158]Fig. 3H, right). Depletion of neutrophils, but not macrophages or dendritic cells, nearly abolished the increased colonization by strong host neutrophilic inflammation ([159]fig. S6, D and [160]E), further validating the pro-metastatic role of neutrophils in vivo. Moreover, such a metastasis-promoting effect of host neutrophilic inflammation was similarly detected in the spontaneous lung metastasis model ([161]fig. S6F). Besides the above tumor-associated neutrophilic inflammation conditions, we further asked whether other types of host inflammation would similarly affect the immunosuppressive capacity of lung neutrophils. To this end, we employed the lipopolysaccharide (LPS)-induced systemic inflammation model ([162]44) ([163]fig. S7A), and found that LPS-induced acute host inflammation not only increased the number of lung-infiltrating neutrophils ([164]fig. S7, B and [165]C), but also reinforced the immunosuppressive capacities of lung neutrophils ([166]fig. S7, D–[167]F), leading to aggravated lung metastatic colonization of breast tumor cells ([168]fig. S7G). Taken together, both tumor-associated neutrophilic inflammation and host systemic inflammation were able to strengthen the immunosuppressive property of lung neutrophils, which consequently heightened breast cancer lung metastasis. Tissue-specific immunosuppressive capacity pertains to human neutrophils To identify whether a tissue-dependent immunosuppressive phenotype and function for neutrophils might similarly occur in human neutrophils, we exploited humanized NSG^™-SGM3 mice ([169]45–[170]47) in which human CD34^+ hematopoietic stem cells have been engrafted and have developed into human immune cells, including human neutrophils marked by CD33^+CD15^+CD66b^+ ([171]31, [172]40, [173]48) ([174]fig. S8A). We first performed scRNA-seq to compare BM-, PB- and lung-derived human CD45^+CD33^+ myeloid cells isolated from naïve humanized mice ([175]Fig. 4A). Among the 14 clusters revealed by unbiased clustering, clusters 0, 2, 3, 7 and 11 were identified as neutrophils based on their expression of the canonical markers S100A8 and S100A9 ([176]17) ([177]Fig. 4B). We then selected these 5 populations to further perform a reclustering analysis and mapped the tissue-specific information onto them ([178]Fig. 4C). Consistent with our findings from mouse neutrophils, human neutrophils differentiated in humanized mice exhibited the same tissue-specific clustering pattern, and in particular, lung neutrophil clusters did not significantly overlap with those of BM or PB neutrophils ([179]Fig. 4C). Fig. 4. Tissue-specific immunosuppressive characteristic pertains to human neutrophils. Fig. 4. [180]Open in a new tab (A) A schematic showing scRNA-seq of human CD45^+CD33^+ myeloid cells isolated from the BM, PB and lung of naïve humanized mouse (n=1, no sample pooling). (B) t-SNE plots showing the distribution of myeloid cells from the data in (A). Neutrophils are indicated according to the markers S100A8 and S100A9. (C) t-SNE plots showing the distribution of human neutrophils. Lung neutrophils are indicated, and plots at right highlight cells from each individual tissue. (D) The expression of indicated genes on the t-SNE plot from the data in (C). (E) As depicted in the schematic ([181]fig. S8C), from the scRNA-seq data of human neutrophils, violin plots showing the expression levels of the indicated genes in human neutrophils from BM, PB, lung and primary tumor of MDA-MB-231 tumor-bearing humanized mouse (n=1, no sample pooling). (F) Violin plots showing the expression levels of the indicated genes in human donor-derived BM, PB and lung neutrophils (datasets from Tabula Sapiens). (G) Functional enrichment analysis of the up-regulated genes in lung metastasis (Met) was performed using Metascape. Statistical differences between groups were calculated by Kruskal-Wallis Test (E-F). *p< 0.05; **p< 0.01; ***p< 0.001; ****p< 0.0001; NS, not significant. Probing the immunosuppression-associated genes, such as PTGS2, IL1B, TREM1, NFE2L2, VEGFA, C5AR1 and OLR1, we found that they were again mainly enriched in lung-infiltrating human neutrophils, while being barely expressed in the BM- or PB-derived human neutrophils ([182]Fig. 4D). Within the two clusters (clusters 0 and 2) of lung neutrophils, those immunosuppression-associated genes did not show a preferential expression in specific clusters ([183]fig. S8B). Under human breast tumor-bearing condition ([184]fig. S8C), the expression levels of immunosuppression-associated genes in lung-infiltrating human neutrophils were again more pronounced than their counterparts isolated from BM and PB, as well as those from primary tumors ([185]Fig. 4E). Therefore, under both steady-state and tumor-bearing conditions, human lung neutrophils showed characteristically high expression of immunosuppression-associated genes. We next leveraged the recently published “Tabula Sapiens” dataset, which profiled the transcriptomes of more than 400 cell types in multiple organs of the human body by scRNA-seq ([186]49). Consistently, lung neutrophils from human donors showed higher expression of the immunosuppression-associated genes, including PTGS2, IL1B, TREM1, NFE2L2, CD14 and CLEC4E, compared to human neutrophils from BM or PB ([187]Fig. 4F). Further, we compared the transcriptomic profiles of lung metastases (Met) to those of metastases to other organs (bone, brain and liver) ([188]GSE14018) ([189]50) or to those of matched primary tumors ([190]GSE110590) ([191]51) from published human breast cancer datasets. Enriched pathway analysis indicated that, compared to other organ metastases and matched primary tumors, lung metastases were enriched in pathways associated with neutrophil migration and activation, and negative regulation of lymphocyte and T cell activation ([192]Fig. 4G and [193]fig. S8D). Collectively, these data suggested that the tissue-specific immunosuppressive nature of lung-infiltrating neutrophils pertains to humans. Lung mesenchymal cells reprogram neutrophils to be immunosuppressive To test whether lung microenvironment contains certain driver stromal cells serving to endow neutrophils with immunosuppressive properties and characterize the specific lung tissue cells involved, we co-cultured mouse BM-derived neutrophils ex vivo with various types of freshly isolated lung stromal cells from naïve mice ([194]Fig. 5A and [195]fig. S9A). Using the expression of a subset of the above-defined immunosuppression-associated genes (Ptgs2, Il1b, Arg2, and Trem1) as an indicator of neutrophil reprogramming, we found that, among the examined lung tissue cells, CD140a^+ (also known as platelet-derived growth factor receptor alpha, PDGFRα) lung mesenchymal cells (MCs) significantly increased neutrophil expression of these genes ([196]Fig. 5B). Moreover, an immature state did not abrogate neutrophils’ capacity to be reprogrammed by MCs ex vivo, although immature neutrophils were likely less modifiable than mature neutrophils in their ability to upregulate immunosuppression-associated genes upon lung MC stimulation ([197]fig. S9B). Consistent with the in vivo tissue environment-dependent neutrophil reprogramming we observed ([198]Fig. 1F), lung MCs were found to be superior to CD140a^+ MCs isolated from other tissues (BM, spleen, and liver) in their ability to induce the immunosuppression-associated genes ex vivo ([199]Fig. 5C). Therefore, MCs were identified as the stromal component in the lung that reprograms neutrophils at the steady-state. Fig. 5. Lung mesenchymal cells reprogram neutrophils to be immunosuppressive. Fig. 5. [200]Open in a new tab (A-B) As depicted in the schematic (A), BM-derived neutrophils were monocultured or co-cultured with the indicated lung tissue cells isolated from naïve mice. Relative mRNA expression of the indicated genes was measured (B) (n=4). (C) BM-derived neutrophils were monocultured or co-cultured with MCs isolated from the indicated tissues. Relative mRNA expression of the indicated genes in neutrophils was measured (n=4). (D) Heatmap showing the expression of the indicated genes from the RNA-seq data of control neutrophils and lung MC-educated neutrophils (n=3). (E) Venn diagram of differentially expressed genes (DEGs) in lung neutrophils (data in [201]fig. S2A) and lung MC-educated neutrophils (data in [202]Fig. 5D). (F) Control BM-derived neutrophils or lung MC-educated neutrophils were co-cultured with splenic CD3^+ T cells at a ratio of 1:1 in the presence of plate-coated anti-CD3 and soluble anti-CD28. T cell proliferation was quantified by flow cytometry (n=4). (G) Immunostaining of lung section showing the localization of MCs (green) and neutrophils (Ly6G, red) in naïve CD140a^EGFP mice. Scale bars, 50 μm. (H) BM-derived neutrophils were monocultured or co-cultured with lung MCs or lung MC-derived conditioned medium. Relative mRNA expression of the indicated genes was measured (n=4). (I) Relative mRNA expression of the indicated genes in BM-derived neutrophils after monocultured or co-cultured with conditioned medium from naïve mice-, AT3 tumor-bearing mice- or AT3-gcsf tumor-bearing mice-derived lung MCs (n=5). n represents the number of biological replicates. The results in B, C, and F-I are representative of three independent experiments and shown as mean ± SEM. Statistical significance was determined by one-way ANOVA with Tukey’s multiple comparisons test (B, C, F, H-I). *p< 0.05; **p< 0.01; ***p< 0.001; ****p< 0.0001; NS, not significant. Through RNA-seq, we next attempted to gain a broader understanding of how lung MCs modulate neutrophils at the transcriptional level. In addition to the four “indicator” genes selected above, the other lung neutrophil-specific immunosuppression-associated genes ([203]Fig. 1C and [204]fig. S2A) were upregulated in BM neutrophils upon their co-culture with lung MCs ([205]Fig. 5D). Apart from these immunosuppression-related genes, the top-ranked lung neutrophil signature genes as identified by scRNA-seq ([206]fig. S4B), such as Cxcl2, G0s2, Ccrl2, Acod1, Il1r2, Cebpb, and Ets2, were also found to be induced in BM neutrophils upon co-culture with lung MCs ([207]fig. S9C). By further comparison of the differentially expressed genes (DEGs) between the datasets derived from mouse lung neutrophils and lung MC-educated BM neutrophils, we found that a large portion of their DEGs overlapped ([208]Fig. 5E). These results clearly indicated that lung MCs broadly remodel BM neutrophil transcriptomes into a lung neutrophil-like phenotype. In alignment with the above transcriptional reprogramming, lung MC-educated neutrophils were found to be functionally strengthened in their suppression of T cell proliferation ([209]Fig. 5F), and NK cell cytotoxicity ([210]fig. S9D). A close physical interaction was detected between CD140a^+ MCs and Ly6G^+ neutrophils, as shown by immunostaining of naïve mouse lung sections ([211]Fig. 5G and [212]fig. S9E), supporting lung MC-mediated neutrophil reprogramming in situ. In human lungs, a similar spatial distribution pattern was observed between CD140A (PDGFRA)-expressing spots (MC-enriched) and CSF3R- or S100A8-expressing spots (neutrophil-enriched) through analysis of published lung spatial transcriptomic data ([213]GSE178361) ([214]52) ([215]fig. S9F). Intriguingly, despite the proximity of lung MCs and neutrophils, cell–cell contact was not necessary for lung MCs to modulate neutrophils, as lung MC-derived conditioned medium was as effective as lung MCs in upregulating immunosuppression-associated genes in neutrophils ([216]Fig. 5H). Hence, lung MCs serve to reprogram neutrophils at the transcriptional and functional levels at the steady-state, likely through soluble factors. In consideration of the reinforcing effects of host inflammation conditions on the immunosuppressive properties of lung neutrophils, we wondered whether these effects were mediated by lung MCs. First, we detected that lung CD140a^+ MCs had elevated proliferation and increased total cell numbers at the pre-metastatic stage of the 4T1 and AT3-gcsf models, both of which are associated with robust neutrophilic inflammation in the host lungs ([217]fig. S9, G and [218]H). Through further in vivo and ex vivo assays, we reasoned that lung neutrophils serve to stimulate lung MCs to express a series of stromal growth factors that may promote MC proliferation in an autocrine manner ([219]fig. S9, I–[220]K). Second, the strength of neutrophil-reprogramming capacity of lung MCs was elevated when MCs were derived from AT3 tumor-bearing mice vs. naive mice, and was further strengthened when using lung MCs from AT3-gcsf tumor-bearing mice ([221]Fig. 5I, [222]fig. S9L and [223]fig. S9M). Consistently, in vivo, exogenously implanted BM neutrophils acquired successively heightened expression of immunosuppression-associated genes from naïve, AT3 tumor-bearing, to AT3-gcsf tumor-bearing conditions ([224]fig. S9N). Taken together, lung MCs have an intrinsic capacity to reprogram infiltrating neutrophils to be immunosuppressive, and this capacity is further reinforced by tumor-associated host inflammation conditions. PGE2 is the key factor by which lung MCs endow neutrophils with an immunosuppressive phenotype As lung MC-mediated neutrophil reprogramming was independent of cell-cell contact, we next sought to identify the possible lung MC-derived soluble factor(s) that functions to modulate neutrophils. Through ingenuity pathway analysis (IPA) of the RNA-seq data from lung neutrophils vs. PB neutrophils ([225]fig. S2A) or lung MC-educated neutrophils vs. control neutrophils ([226]Fig. 5D), there emerged a number of predicted upstream regulators ([227]Fig. 6, A and [228]B) that may induce the changes in neutrophils. Of note, overlapping top upstream candidate regulators from both datasets included tumor necrosis factor alpha (TNFα), interleukin 1 beta (IL-1β), interferon gamma (IFNγ), prostaglandin E2 (PGE2), interleukin 4 (IL-4), interleukin 33 (IL-33), interleukin 17A (IL-17A) and interleukin 6 (IL-6), which have known roles in immune regulation. Through an in vitro screen of these factors, PGE2, an inflammatory lipid mediator, was superior in upregulating immunosuppression-associated genes in BM neutrophils ([229]Fig. 6C). Further supporting our proposed signaling model, lung MCs were the main producers of PGE2 vs. other types of lung tissue cells, as they expressed a higher level of Ptgs2, which encodes the rate-limiting enzyme in PGE2 synthesis, and produced more abundant PGE2 ([230]fig. S10A). Abrogation of PGE2 production using Ptgs2^−/− lung MCs ([231]fig. S10B) largely reduced MC-stimulated neutrophil expression of immunosuppression-associated genes ([232]Fig. 6D), and reversed T cell and NK cell suppression by lung MC-educated neutrophils ([233]fig. S10, C and [234]D). Therefore, PGE2 serves as a lung MC-derived soluble factor that reprograms neutrophils. Fig. 6. PGE2 is the key factor by which lung MCs endow neutrophils with an immunosuppressive phenotype. Fig. 6. [235]Open in a new tab (A-B) Ingenuity pathway analysis was performed to predict the upstream regulators on the basis of differentially expressed genes from lung neutrophils vs. PB neutrophils (A), or MC-educated neutrophils vs. control (Con) neutrophils (B). The overlapping candidate regulators are indicated in red font. (C) The mRNA expression of the indicated genes in BM-derived neutrophils was analyzed after stimulation with the indicated regulators (n=4). (D) BM-derived neutrophils were monocultured or co-cultured with lung MCs isolated from WT or Ptgs2^−/− mice and relative mRNA expression of the indicated genes in neutrophils was determined (n=4). (E) BM-derived neutrophils were stimulated with IL-1β, IL-6, PGE2 or their combinations. Relative mRNA expression of the indicated genes in neutrophils was determined (n=4). (F) Heatmap showing expression of the indicated genes from the RNA-seq data of control neutrophils and IL-1β/IL-6/PGE2-treated neutrophils (n=3). (G) Venn diagram of differentially expressed genes (DEGs) in lung MC-educated neutrophils (data in [236]Fig. 5D) and IL-1β/IL-6/PGE2-treated neutrophils (data in [237]Fig. 6F). (H-I) Neutrophils were pre-treated with vehicle or IL-1β/IL-6/PGE2 and then their immunosuppressive effects on T cells (H) or NK cells (I) were measured (n=3–5). Neutrophil: T or NK cell = 1:1. n represents the number of biological replicates. The results in C-E, and H-I are representative of three independent experiments and shown as mean ± SEM. Statistical significance was determined by one-way ANOVA with Tukey’s multiple comparisons test (C-E, H-I). *p< 0.05; **p< 0.01; ***p< 0.001; ****p< 0.0001; NS, not significant. We next investigated how PGE2 in lung MCs was regulated under tumor-bearing conditions. In the AT3 model associated with weak inflammation, the level of lung resident PGE2 underwent a mild increase compared to naïve mice, while it was strikingly elevated in the AT3-gcsf and 4T1 models ([238]fig. S10, E and [239]F). Correspondingly, lung MCs derived from AT3-gcsf tumor-bearing mice were found to produce more abundant PGE2 than those from AT3 tumor-bearing or naïve mice ([240]fig. S10G). Further ex vivo experiments demonstrated that G-CSF-expanded lung neutrophils, but not G-CSF itself, acted as the stimulator to elicit PGE2 production in lung MCs ([241]fig. S10, H and [242]I). In addition, epidermal growth factor (EGF), which was elevated in the lung environment under the weak inflammatory condition (AT3 model), was also identified in a screen to be capable of stimulating lung MCs to produce PGE2 ([243]fig. S10, J–[244]L). In addition to PGE2, among the top upstream candidate regulators ([245]Fig. 6, A and [246]B), IL-1β and IL-6 are two inflammatory cytokines known to play critical roles in MDSC development and activation ([247]20). Further ex vivo experiments showed that IL-1β and IL-6 were indeed capable of enhancing the effects of PGE2 in upregulating neutrophil immunosuppression-associated genes ([248]Fig. 6E). In vivo, neutralization of IL-1β and IL-6 significantly reduced expression of immunosuppression-associated genes in lung neutrophils at the steady-state ([249]fig. S10M). Lung MCs and neutrophils were identified as the main producers of IL-6 and IL-1β, respectively ([250]fig. S10, N–[251]P). Under tumor-associated host inflammation conditions, secretion of IL-6 and IL-1β was further elevated ([252]fig. S10, Q and [253]R). Hence, PGE2, IL-1β, and IL-6 were characterized as the main lung environmental factors acting together to elicit neutrophil reprogramming. A combination of PGE2, IL-1β and IL-6 acted similarly to lung MCs in immunosuppressive reprogramming of neutrophils at the transcriptional level ([254]Fig. 6, F and [255]G) and functional level ([256]Fig. 6, H and [257]I). Collectively, through secretion of soluble PGE2, lung MCs endow infiltrating neutrophils with an acquired immunosuppressive capacity, while MC-derived IL-6 and neutrophil-derived IL-1β further enhance this effect. Identification of putative cell-cell communications between MCs and neutrophils To further delineate the molecular profile of PGE2-producing MCs and identify potential MC-neutrophil crosstalk, especially under host neutrophilic inflammation conditions, we performed scRNA-seq on CD140a^+ lung MCs isolated from the pre-metastatic stage of AT3-gcsf tumor-bearing mice ([258]Fig. 7A). Unbiased clustering ([259]53) identified 8 clusters within the CD140a^+ lung MCs ([260]Fig. 7B), and clusters 0 and 4 were found to express remarkably higher levels of Ptgs2 than other clusters ([261]Fig. 7, C and [262]D). Through analyzing the top-ranked marker genes in each MC cluster ([263]fig. S11), we found that the Ptgs2^high CD140a^+ MC subsets (clusters 0 and 4) highly expressed several inflammation-associated cytokines and chemokines ([264]Fig. 7D), including Il6, Ccl11, Ccl7, Cxcl1, and Ccl2, which have been reported to regulate myeloid cell recruitment and activation ([265]54, [266]55). These results prompted us to further characterize potential new ligand-receptor interactions between Ptgs2^high MCs and neutrophils, with the goal of identifying targetable interactions to renormalize lung immunity. Fig. 7. Identification of putative cell-cell communications between lung MCs and neutrophils. Fig. 7. [267]Open in a new tab (A). A schematic showing scRNA-seq analysis of CD45^− CD31^− CD326^− EGFP^+ cells that were isolated from AT3-gcsf tumor-bearing CD140a^EGFP mouse (n=1, no sample pooling). (B-C) t-SNE plots (B) and feature plots (C) showing the Ptgs2^high MCs (clusters 0 and 4) among lung CD140a^+ MCs from AT3-gcsf tumor-bearing CD140a^EGFP mouse. (D) Violin plots showing the expression of the indicated genes across each cluster. (E) The total interaction strength from each group was visualized by circle plot. The labeled numbers in the color edge indicated the quantified interaction strength between two cell groups. (F) The dominant signaling senders (sources) and signaling receivers (targets) were visualized in a 2-dimensional space using a scatter plot. (G) Heatmap showing the signaling pathways that contributed to the outgoing (left) and incoming signaling (right) of certain cell groups. (H) The significant interactions (ligand-receptor pairs) associated with the indicated pathways from MC groups (M0 and M4) to neutrophil groups (N1, N3 and N6) were visualized by bubble plot. (I) Dot plots showing the expression of the indicated genes across each MC cluster. (J) The significant interactions (ligand-receptor pairs) associated with ‘CXCL’ and ‘CCL’ signaling pathways from MC groups (M0 and M4) to neutrophil groups (N1, N3 and N6) were visualized by chord diagram. (K) The signaling pathways from neutrophil groups (N1, N3 and N6) to MC groups (M0 and M4) were visualized by chord diagram. (L) The significant interactions (ligand-receptor pairs) associated with ‘SELL’ and ‘ITGAL-ITGB2’ signaling pathways from neutrophil groups (N1, N3 and N6) to MC groups (M0 and M4) were visualized by bubble plot. (M) Violin plots showing the expression of the indicated genes associated with ‘SELL’ and ‘ITGAL-ITGB2’ signaling pathways across each cell group. To this end, we proceeded with cell-cell communication analysis of the above two Ptgs2^high MC clusters and the lung neutrophil clusters (clusters 1, 3, and 6, [268]Fig. 3A, right), all of which were derived from the AT3-gcsf model, by CellChat ([269]56). There were 185 significant ligand-receptor interactions identified pairwise amongst the five cell groups, which were further categorized into 54 signaling pathways. Through examining the signaling strengths derived from each cell cluster, the two lung MC groups (M0 and M4) were found to be the primary signaling sources compared to the three neutrophil groups (N1, N3, and N6) ([270]Fig. 7E), suggesting that MCs may play a more dominant role in MC-neutrophil communications. In addition to the reciprocal paracrine interactions detected between MC and neutrophil clusters, considerable cellular interactions were also found to occur within each cluster in an autocrine manner ([271]Fig. 7E). Upon distinguishing the signaling senders (sources) and signaling receivers (targets), we found that the two MC groups were the primary sources of outgoing signaling, while the three neutrophil groups mainly serve to receive the signaling ([272]Fig. 7F). Furthermore, we ranked the 54 signaling pathways based on their contribution to the outgoing and incoming signaling of the communication networks and identified the top 10 signaling pathways ([273]Fig. 7G). Among them, except for the ‘SELPLG’ pathway, which was a completely autocrine neutrophil signal, the other nine signaling pathways -- ‘COLLAGEN’, ‘CXCL’, ‘LAMININ’, ‘ANNEXIN’, ‘CCL’, ‘FN1’, ‘THBS’, ‘COMPLEMENT’, and ‘ICAM’ -- were involved in MC-neutrophil communications ([274]Fig. 7G). Through a further analysis of the significant ligand-receptor pairs that contributed to the signaling from MCs to neutrophils within these nine signaling pathways ([275]Fig. 7H), we identified 25 ligands and 14 receptors in MCs and neutrophils, respectively. Col1a1 and Col1a2 were the major ligands contributing to ‘COLLAGEN’ signaling, which are known MC signature genes ([276]57). Besides that, C3, Ccl7, Cxcl1 and Fn1 were found to be dominant ligands for outgoing signaling of MCs ([277]Fig. 7H). For neutrophils, Cd44, the gene encoding a cell-surface glycoprotein involved in neutrophil recruitment and activation ([278]58, [279]59), was identified as the primary receptor in multiple pathways including ‘COLLAGEN’, ‘LAMININ’ and ‘FN1’. In addition to Cd44, Cxcr2, Ccr1, and some integrins also contributed to the signaling reception ([280]Fig. 7H). By mapping the expression of above ligands in CD140a^+ lung MCs ([281]Fig. 7B), we found that most of the ligands were not specific for the M0 and M4 clusters, but also expressed in other MC clusters ([282]Fig. 7I), suggesting that these defined intercellular interactions ([283]Fig. 7H) may generally occur between lung CD140a^+ MCs and neutrophils. Notably, Cxcl1 and Ccl7, two signature genes of the Ptgs2^high MC populations ([284]Fig. 7D and [285]fig. S11), were characterized as the dominant sources for the signaling from MCs to neutrophils in ‘CXCL’ and ‘CCL’ pathways, respectively ([286]Fig. 7J). Moreover, upon similar analysis of the signaling from neutrophils to MCs, we identified the integrin and adhesion molecule-associated pathways ‘SELL’ and ‘ITGAL-ITGB2’, and in particular the ligand-receptor pairs Sell-Cd34 and Itgb2-Icam1, as the major contributors ([287]Fig. 7, [288]K–[289]M). Taken together, these cell-cell communication analyses suggested that MCs and neutrophils have a bidirectional crosstalk via multiple ligand-receptor interactions, in which collagens, chemokines, integrins, and adhesion molecules may play more critical roles. Through these interactions, neutrophils are expected to be recruited and engaged within the MC-enriched lung niches, followed by PGE2-mediated immunosuppressive reprogramming. Genetic or pharmacological inhibition of PGE2 signaling reduces the immunosuppressive capacity of lung neutrophils and mitigates lung metastasis of breast cancer To better understand the in vivo role of the PGE2-dependent lung MC program in reprogramming neutrophils, we generated MC-targeted Ptgs2 conditional KO (cKO) mice (Pdgfra-Cre; Ptgs2^flox/flox, or Ptgs2^ΔMCs). We then determined how Ptgs2/PGE2 deficiency in MCs ([290]fig. S12A) alters the immunosuppressive capacity of lung neutrophils in vivo. At the steady-state, although the total number of lung neutrophils remained unchanged in Ptgs2^ΔMCs mice ([291]fig. S12B), their expression of immunosuppression-associated genes was notably lower than those in WT lung neutrophils ([292]Fig. 8A). Accordingly, lung neutrophils isolated from Ptgs2^ΔMCs mice underwent a significant reduction in their suppression of T cell proliferation and NK cell cytotoxicity, compared to their WT counterparts ([293]Fig. 8, B and [294]C). In the modified experimental metastasis model that the functions of pre-metastatic niches can be gauged ([295]14, [296]43), Ptgs2^ΔMCs mice developed a lower level of lung metastasis than their WT littermates, while the primary tumor growth was not altered by Ptgs2 cKO ([297]Fig. S12C). This was similarly detected ([298]fig. S12D) using another syngeneic breast tumor model, E0771, which is genetically distinct from the AT3 cell line ([299]60). Therefore, MC-specific ablation of Ptgs2/PGE2 led to an impairment of the immunosuppressive capacity of lung neutrophil and mitigated lung metastasis of breast cancer. Fig. 8. Targeting PGE2 signaling reduces the immunosuppressive capacity of lung neutrophils and improves the therapeutic efficacy of adoptive T cell-based immunotherapy in treating lung metastasis. Fig. 8. [300]Open in a new tab (A) Relative mRNA expression of the indicated genes was measured in lung neutrophils isolated from WT or Ptgs2^ΔMCs naïve mice (n=6). (B-C) Lung neutrophils were isolated from WT or Ptgs2^ΔMCs naïve mice, and then their immunosuppressive effects on T cells (B) or NK cells (C) were determined (n=4). Neutrophil: T or NK cell = 1:1. (D) BM-derived neutrophils were monocultured or co-cultured with lung MCs in the presence of different blocking agents. Relative mRNA expression of the indicated genes was measured in neutrophils (n=4). (E) As depicted in the schematic (left, up), immunostaining results of lung section showing the localization of implanted T cells (green) and neutrophils (Ly6G, red) in AT3 tumor-bearing and AT3-gcsf tumor-bearing mice (right). Scale bars, 50 μm. The distance of implanted T cell to nearest neutrophil was measured (left, bottom). (F-G) As depicted in the schematic (F, left), total number of implanted OT-I CD8^+ T cells (F, right) and the frequency of IFNγ^+ OT-I CD8^+ T cells (G) were measured in AT3 tumor-bearing and AT3-gcsf tumor-bearing mice (n=7). (H) As depicted in the schematic (left), the therapeutic efficacies of adoptively transferred OT-I CD8^+ T cells in treating lung metastasis under the AT3 tumor-bearing and AT3-gcsf tumor-bearing conditions were determined by ex vivo BLI (n=9). (I) As depicted in the schematic (left), the combined effect of dual inhibition of EP2 and EP4 and adoptively transferred OT-I CD8^+ T cells in treating lung metastasis under the AT3-gcsf tumor-bearing condition was determined by ex vivo BLI (n=10–12). n represents the number of biological replicates. The results are representative of two independent experiments (A-I) and shown as mean ± SEM. Statistical significance was determined by unpaired two-tailed Student’s t-test (A, E-G), one-way ANOVA with Tukey’s multiple comparisons test (B-D), unpaired two-tailed Mann-Whitney test (H), or Kruskal-Wallis test (I). *p< 0.05; **p< 0.01; ***p< 0.001; ****p< 0.0001; NS, not significant. Next, we attempted to pharmacologically block the PTGS2/PGE2 signaling pathway as an endeavor to develop clinically applicable approaches to treat lung metastasis. As we recently reported, however, an FDA-approved PTGS2 (COX-2) inhibitor celecoxib failed to reduce lung tissue resident PGE2 level and alleviate lung metastasis ([301]43). We then reasoned that the ineffectiveness of PTGS2 inhibition, which was similarly reported in recent breast cancer clinical trials ([302]61–[303]64), might be due to a compensatory effect from COX-1, another isoform of cyclooxygenase (PTGS) ([304]65). We therefore speculated that blockade of PGE2 receptors might circumvent any compensatory effects of PGE2-producing enzymes and would thus represent a more effective avenue to inhibit PGE2 signaling in vivo. From the scRNA-seq results of naïve mouse lung neutrophils ([305]fig. S12E), we found that among the four PGE2 receptors (EP1-EP4), lung neutrophils expressed relatively high levels of EP2 (encoded by Ptger1) and EP4 (encoded by Ptger4), which was further confirmed by qPCR ([306]fig. S12F). Taking into account possible compensatory effects between EP2 and EP4 downstream signaling ([307]66, [308]67), we investigated whether targeting both receptors simultaneously would be more effective than blocking either one alone. In the ex vivo MC-neutrophil co-culture system, dual inhibition of EP2 and EP4 indeed showed a superior effect in diminishing lung MC-induced immunosuppression-associated gene expression in neutrophils compared with blockade of either receptor alone, or blockade of IL-1β or IL-6 ([309]Fig. 8D). In vivo, administration of EP2 and EP4 antagonists together in naïve mice significantly reduced immunosuppression-associated gene expression in lung neutrophils ([310]fig. S12G). Based on these results, we next evaluated dual inhibition of EP2 and EP4 in controlling lung metastasis in mouse models of breast cancer. In the modified experimental metastasis model, dual inhibition of EP2 and EP4 showed a marked reduction of AT3 cell colonization of the lung under the host neutrophilic inflammation condition ([311]Fig. S12H). Depletion of CD8^+ T cells and NK cells in the above model or employing the same model in immunodeficient NSG mice, which lack T cells and NK cells, largely reduced the metastasis-mitigating effect of EP2 and EP4 antagonists ([312]fig. S12, I–[313]L), suggesting that the neutrophil-mediated T cell and NK cell suppression functionally contributes to the effects of EP2 and EP4 inhibition. Moreover, this metastasis-repressing effect of dual inhibition of EP2 and EP4 was verified using the E0771 cell line ([314]fig. S12M). These results strongly supported that inhibition of both EP2 and EP4 receptors could be a promising strategy in the management of lung metastatic disease. Taken together, blockade of PGE2 signaling through genetic or pharmacological approaches was an effective approach to reduce immunosuppression by lung neutrophils and to restrain breast cancer lung metastasis under host inflammation conditions. Targeting PGE2 signaling improves the therapeutic efficacy of adoptive T cell-based immunotherapy in treating lung metastasis Given the robust capacity of lung neutrophils to suppress T cells described above, particularly under inflammation conditions, we asked whether such an immunosuppressive lung environment may dampen the efficacy of T cell-based therapy and whether blockade of PGE2 signaling may serve as a possible solution to this problem. When fluorescently-labeled CD8^+ T cells were transplanted into recipient mice bearing AT3 or AT3-gcsf orthotopic tumors, a portion of them were spatially close to endogenous neutrophils, which was more prominent in AT3-gcsf tumor-bearing mouse lungs ([315]Fig. 8E). Additionally, the total number of exogenous CD8^+ T cells was significantly lower under strong inflammation conditions (AT3-gcsf tumor-bearing) than under weak inflammation conditions (AT3 tumor-bearing) ([316]Fig. 8F and [317]fig. S13A). Heightened host inflammation also notably reduced the expression of IFN-γ in exogenous CD8^+ T cells in recipient mouse lungs ([318]Fig. 8G and [319]fig. S13B). Consequently, the therapeutic benefit of adoptively transferred CD8^+ T cells in treating lung metastases under AT3 tumor-bearing conditions was not detected under AT3-gcsf-bearing conditions ([320]Fig. 8H). Thus, elevated host inflammation, characterized by the increased immunosuppressive capacity of lung neutrophils, blunted anti-tumor immunity in the lung and was associated with loss of therapeutic efficacy of adoptive T cell-based immunotherapy in treating lung metastases. Next, we investigated whether dual inhibition of EP2 and EP4 could improve the therapeutic efficacy of adoptive T cell transfer therapy. To this end, ex vivo-activated OT-I CD8^+ T cells and EP2/EP4 antagonists, singly or in combination, were administered to AT3-gcsf tumor-bearing mice in the modified experimental lung metastasis model ([321]Fig. 8I, left). Indeed, EP2 and EP4 dual inhibition improved the therapeutic effectiveness of adoptively transferred T cells, as reflected by a >10-fold reduction in lung metastases compared to T cell treatment alone ([322]Fig. 8I). Consistent with the beneficial effect from blockade of PGE2 receptors, MC-specific host Ptgs2 cKO also enabled therapeutic effectiveness of CD8^+ T cell-based therapy in treating established lung metastases ([323]fig. S13C). Therefore, inhibition of the Ptgs2-PGE2-EP2/EP4 signaling pathway was revealed to be a promising combination therapy to synergize with adoptive T cell-based immunotherapy in treating lung metastasis. Altogether, our results identify a mesenchymal cell-neutrophil interacting program and characterize how this program contributes to lung metastasis in breast cancer models ([324]fig. S13D). Our findings provide insights into organ stroma-guided acquired immunosuppression by neutrophils and establish a foundation for the development of new strategies targeting stroma-myeloid cell interactions to prevent and treat metastatic disease. DISCUSSION The roles of organ-infiltrating neutrophils in regulating distant organ metastasis of solid cancers have received increasing attention in recent years ([325]10, [326]11, [327]14, [328]68, [329]69). Prior to tumor metastasis, de novo expanded neutrophils egress from hematopoietic sites, circulate through the bloodstream and infiltrate organs. During this transmitting process, neutrophils undergo phenotypic and functional changes in response to tissue microenvironments ([330]35, [331]70, [332]71). However, how tissue-specific neutrophil properties are functionally associated with the metastatic process remains largely unknown. Our comprehensive analysis of BM, PB, and lung neutrophils across different conditions sheds light on this question. Since neutrophilic inflammation is a hallmark of various lung pathologies ([333]72–[334]75), we surmise that the lung-restricted immunosuppressive property of neutrophils we uncovered may also be relevant to a broad range of lung diseases beyond cancer metastasis, such as lung infections, interstitial lung disease, and primary lung cancer. Lung MCs, a type of lung-resident stromal cells, play essential roles in a variety of lung diseases including asthma, lung fibrosis, acute lung injury, and chronic obstructive pulmonary disease ([335]76–[336]82), but remain poorly characterized for their contribution to lung pre-metastatic niche formation. Our studies identified lung MCs as the major lung stromal cells driving the immunosuppressive reprogramming of neutrophils. Such a capacity can be further reinforced by neutrophilic inflammation or host systemic inflammation (such as LPS-induced inflammation). RNA-seq data from MC-educated neutrophils indicated that lung MCs account for a wide range of tissue-specific features of lung-infiltrating neutrophils in addition to the immunosuppressive properties. Indeed, lung neutrophils have been reported to exert non-immune physiological effects to maintain organ homeostasis, such as support of lung angiogenesis and the regulation of lung circadian rhythms ([337]35, [338]71). Thus, it will be intriguing to further elucidate the roles of lung MCs in endowing tissue-specific capacities to neutrophils, as well as other types of myeloid cells, which could be helpful in identifying new therapeutic targets in related pathological conditions. A recent study revealed how neutrophils tailor their tissue-specific properties to support organ homeostasis and suggested that tissue-derived signals could drive rapid neutrophil adaptation ([339]35). Here, we identified that PGE2 is the key lung-specific signal that endows infiltrated neutrophils with immunosuppressive capacities. Although PGE2 has been widely studied and recognized as an important player in inflammation and cancer ([340]83, [341]84), its roles in the preservation of lung immune homeostasis and formation of the pre-metastatic niche remain less understood. Our work adds another layer of complexity to the biology of PGE2 in cancer metastasis by identifying a role in reprogramming organ-specific neutrophils, which in turn modulate lung-resident immunity. Notably, we demonstrated that the PGE2-EP2/EP4 signaling axis drives neutrophil reprogramming, however, the specific downstream molecules in neutrophils that regulate immunosuppressive gene expression remain to be determined. Among the top-listed signature genes of lung neutrophils, there were some encoding transcription factors that have been reported to regulate the suppressive function of MDSCs, such as Cebpb and Nfkbia ([342]20, [343]85, [344]86). Future efforts could focus on investigating whether these transcription factors signal downstream of PGE2 receptors in neutrophils. Adoptive cell therapy (ACT) has emerged as a promising therapeutic approach in treatment of several human malignancies ([345]87, [346]88). However, one of the biggest challenges that limits the effectiveness of immune cell-based ACT is the presence of an immunosuppressive microenvironment in primary tumors or distant organs that hinders the effector function of adoptively transferred cells ([347]89, [348]90). In this study, we showed that the accumulation of lung immunosuppressive neutrophils dampens the anti-tumor function of adoptively transferred T cells, causing therapeutic inefficacy. Targeting PGE2 signaling, which abrogated immunosuppressive neutrophil induction, was strikingly effective in improving adoptive T cell-based therapy in treating lung metastasis of breast cancer in mouse models. Thus, our findings may facilitate the combination of PGE2 signaling blockade with immunotherapeutics into clinical application to treat metastatic disease. MATERIALS AND METHODS Study design The aim of this study is to investigate how lung-infiltrating neutrophils acquire metastasis-modulating effects. We performed bulk RNA-seq, scRNA-seq, quantitative PCR, confocal microscopy and flow cytometry to comprehensively compare the properties of neutrophils isolated from BM, PB and lungs. We further defined how lung stromal cells reprogram the infiltrated neutrophils to endow them with tissue-specific features. Using mouse models, we evaluated the effects of targeting the stroma-neutrophil regulatory program in controlling lung metastasis of breast cancer. The sample size in each experiment was determined based on the level of expected heterogeneity of the samples, the expected or observed difference, and previous publications and our pilot studies. Unless otherwise indicated, age-matched mice were used for experiments and mice were randomly assigned to different experimental groups. Experimental endpoints were established before any in vivo experiments for animal welfare considerations. The studies were unblinded as analyses were performed using quantifiable parameters to minimize bias. The details of the biological replicates in each group and number of repetitions per experiment are indicated in the respective figure legends. No outliers were excluded. Animals 8- to 12-week-old female mice were used in this study. The mice were maintained under 12-hour light/dark cycles at 22 °C and 40–50% humidity. The mice were fed on a chow diet ad libitum and housed in a specific pathogen-free facility. All animal experiments and related protocols were approved by and conducted in accordance with guidelines from the Institutional Animal Care and Use Committee of The Jackson Laboratory (protocol [349]AUS17027). For more details, see the [350]Supplementary Materials. Cell lines and primary cultures The mouse cell lines (4T1 and YAC-1) and human cell line MDA-MB-231 were purchased from the American Type Culture Collection (ATCC). The mouse cell line E0771 was purchased from CH3 Biosystems and mouse cell line AT3 was kindly provided by Dr. Scott I. Abrams (Roswell Park Comprehensive Cancer Center). For more details, see the [351]Supplementary Materials. Tissue processing For PB cells, whole blood was collected via heart puncture with ethylenediaminetetraacetic acid (EDTA) as the anticoagulant. For BM cells, the ends of the tibia and femur were first cut and then the BM was flushed with RPMI-1640 medium using a 27-gauge needle attached to a 10-ml syringe. Lung, spleen, liver and primary tumors were harvested from euthanized mice. For more details, see the [352]Supplementary Materials. Determination of the pre-metastatic stages of breast cancer models The pre-metastatic stages of the breast cancer models were determined as previously described ([353]13, [354]43). For more details, see the [355]Supplementary Materials. Flow cytometry and cell sorting Single cell suspensions were centrifuged and re-suspended in FACS buffer (PBS supplemented with 2% fetal bovine serum) and stained with panels of antibodies (1:200 dilution) for 30 min on ice. Then the cells were washed and 4,6-diamidino-2-phenylindole (DAPI) was added to indicate dead cells. For the intracellular cytokine staining of IFNγ in implanted T cells, lung single cell suspensions were prepared from AT3 and AT3-gcsf tumor bearing mice. Neutrophils were first depleted using the Ly-6G microbeads to avoid the ROS interference ([356]91), prior to the intracellular cytokine staining procedure. Antibody information is shown in [357]table S1. For more details, see the [358]Supplementary Materials. RNA sequencing For the RNA-seq of different tissue-derived neutrophils, neutrophils were freshly sorted from 4T1 or AT3 tumor-bearing mice at the pre-metastatic stage. For the RNA-seq of ex vivo cultured neutrophils, BM-derived neutrophils were co-cultured with lung MCs or stimulated with IL-1β/IL-6/PGE2. Total RNA extraction, sample quality check and the libraries preparation were performed by the Genome Technologies core facility at The Jackson Laboratory. For more details, see the [359]Supplementary Materials. Sample preparation for single-cell RNA sequencing BM, PB and lung neutrophils were sorted from naïve or tumor-bearing mice (4T1, AT3 or AT3-gcsf orthotopic model). Human CD33^+ myeloid cells were sorted from BM, PB and lungs of naïve humanized mice, or BM, PB, lungs and primary tumors of MDA-MB-231 tumor-bearing humanized mice. Lung CD140a^+ MCs were sorted from AT3-gcsf tumor-bearing CD140a^EGFP mice. For more details, see the [360]Supplementary Materials. Single-cell RNA sequencing analysis The Seurat software package (version 4.0.2) in R (version 4.0.2) was used for the downstream analysis of the scRNA-Seq data. The datasets were demultiplex, and the doublet and unmapped cells were removed. Then the singlet cells were filtered to remove low quality and dying cells and used for further analysis. For more details, see the [361]Supplementary Materials. Publicly available datasets The RNA-seq data of neutrophils from germ-free (GF) and specific-pathogen free (SPF) mice are available at the Gene Expression Omnibus (GEO) repository under accession number [362]GSE142432 ([363]35). The RNA-seq data of BM, PB and lung neutrophils from 4T1 tumor-bearing mice are available at the ArrayExpress under accession number E-MTAB-9128 ([364]14). scRNA-seq data of human donor-derived BM, PB and lung neutrophils are from The Tabula Sapiens ([365]49) and available through [366]https://tabula-sapiens-portal.ds.czbiohub.org/. The human RNA-seq data are downloaded from [367]GSE110590 ([368]51) and [369]GSE14018 ([370]50). The spatial transcriptomic data of human lungs are downloaded from [371]GSE178361 ([372]52). Pathway enrichment analysis We employed the Metascape for the functional enrichment analyses. Briefly, the up-regulated genes in lung metastasis vs. other organ metastasis (GEO: [373]GSE14018) or lung metastasis vs. matched primary tumors (GEO: [374]GSE110590) were calculated, and the top 200 genes of each comparison were selected as the inputs for enrichment analysis with default parameters. Only the biological process (BP) of Gene Ontology (GO) with a P value < 0.01 and annotated to ≥3 genes was considered significant. T cell immunosuppression assay Spleens were dissected from naïve mice and single cell suspension was prepared as mentioned above. Then CD3^+ T cells were purified and labeled with proliferation dye CellTrace^™ CFSE (Thermo Fisher Scientific). For in vivo studies, neutrophils were freshly isolated from different tissues and used for experiments. For in vitro studies, BM-derived neutrophils were pre-educated with lung MCs or IL-1β/IL-6/PGE2. To perform this assay, neutrophils were co-cultured with T cells at the ratio of 1:1 in the presence of plate-bound anti-mouse CD3 and soluble anti-CD28. After 2 days of culture, cells were collected and stained with indicated panels of antibodies. T cell proliferation was determined by flow cytometry. NK cell cytotoxicity assay Neutrophils were first isolated from different tissues or pre-treated with lung MCs or IL-1β/IL-6/PGE2. Then the neutrophils were co-cultured with naïve mice-derived splenic NK cells at the ratio of 1:1 in a 96-well round bottom plate at 37°C. Two hours later, the fluorescently labeled target tumor cells were added and co-cultured for another 4 hours. The percentage of tumor cell killing was indicated by propidium iodide (PI) staining and quantified by flow cytometry. Immunofluorescence To detect the immunosuppressive factors in neutrophils, BM, PB and lung neutrophils were freshly isolated from 4T1 tumor-bearing mice at the pre-metastatic stage. To detect the in situ interaction between lung neutrophils and T cells, the distribution of neutrophils in different organs (lung, liver, brain, heart, and kidney), and the distance of T cell to nearest neutrophil or CD11c^+ cell, the mice were euthanized, and indicated tissues were harvested and immediately fixed with 4% paraformaldehyde (PFA) at 4°C. To detect the localization of lung CD140a^+ MCs and neutrophils in the steady state, the lungs were harvested from CD140a^EGFP naive mice and fixed. To detect the localization of in situ neutrophils and implanted OT-I CD8^+ T cells, the mice were first IV injected with fluorescently labeled T cells, and then the lungs were harvested and fixed. For more details, see the [375]Supplementary Materials. Cytospin and Wright-Giemsa staining BM-derived CD101^− and CD101^+ neutrophils were isolated from naïve mice, and then spun onto glass slides (1 × 10^5 cells each) by Cytospin. The slides were fixed in methanol and stained with Wright-Giemsa (Sigma-Aldrich) according to the manufacturer’s instructions. Images were acquired with Leica DM6 with a 60× oil immersion objective, and the brightness was adjusted with ImageJ (1.53c). RNA extraction and quantitative PCR RNA extraction was performed using Direct-zol RNA Miniprep Plus kit (Zymo Research) following the manufacturer’s instructions. cDNA was generated with High-Capacity cDNA Reverse Transcription kit (Thermo Fisher Scientific). RT-qPCR was performed using PowerUp SYBR^™ Green PCR Master Mix (Thermo Fisher Scientific) in 384-well plates and conducted with the ViiA 7 Real-Time PCR system (Thermo Fisher Scientific). 2^−ΔΔCt method was used to calculate the relative gene expression level, with Rps18 as the housekeeping gene. Detailed primer information is shown in [376]Table S2. Experimental and spontaneous lung metastasis models The experimental lung metastasis model was performed to determine the role of pre-metastatic lung niche in regulating tumor cell colonization, as previously described ([377]14, [378]43), with minor modifications. Briefly, to compare breast tumor cell colonization in AT3 and AT3-gcsf tumor-bearing conditions, naïve C57BL/6J mice were first orthotopically implanted with AT3 or AT3-gcsf tumor cells (2 × 10^5 cells/each mouse). On day 12, AT3-Luc tumor cells were intravenously (IV) injected into tumor-bearing mice (1 × 10^6 cells/each mouse). The mice were euthanized on day 26 and lung tissues were rapidly excised for ex vivo BLI. For details in other similar experimental lung metastasis settings, see the [379]Supplementary Materials. To compare the spontaneous lung metastasis progression in AT3 and AT3-gcsf tumor-bearing mice, naïve mice were orthotopically implanted with AT3-Luc or AT3-gcsf-Luc tumor cells (2 × 10^5 cells/each mouse). On day 30, mice were euthanized and the lung tissues were rapidly excised for ex vivo BLI. The ex vivo BLI was performed by putting lung tissues in a 24-well plate containing 2 ml D-luciferin solution (150 μg/ml; diluted in PBS), which were then imaged with the Xenogen IVIS system. Light emission from the region of interest was quantified as photons/second/cm^2/steradian (p/sec/cm^2/sr) using the Living Images software. Plerixafor-mediated neutrophil mobilization To investigate whether the reprogramming of neutrophils is reversible, naïve mice were injected with plerixafor (5 mg/kg). At different time points, the cell number of blood neutrophils was quantified with count beads by flow cytometry according to the manufacturer’s instructions and the immunosuppression-associated gene expression in blood neutrophils was measured by qPCR. LPS model To determine whether LPS-induced acute host inflammation would affect the immunosuppressive capacity of lung neutrophils, naïve mice were injected with vehicle or LPS (2.5mg/kg), and 16 hours later, lungs were harvested and used for further analyses. ELISA and cytokine array ELISA and cytokine array were performed with commercial kits following the manufacturer’s instructions. For more details, see the [380]Supplementary Materials. Illustration tool The schematics are created with [381]BioRender.com. Statistical analysis The details of the biological replicates and repetitions for each experiment were listed in the respective figure legends. In all experiments, results were shown as mean ± SEM (standard error of the mean) to indicate the variation within each experiment. Statistical analyses were performed with GraphPad Prism software (version 8.2.1). Comparison of two groups was done by unpaired two-tailed Student’s t-test or Mann-Whitney test, as indicated in the figure legends. Multiple groups were compared by one-way analysis of variance (ANOVA) or two-way ANOVA, followed by multiple comparison test, as indicated in the figure legends. Statistical significance is defined as * p < 0.05; ** p < 0.01; *** p < 0.001; **** p < 0.0001; or NS (not significant, p > 0.05). Supplementary Material Supplemental Materials [382]NIHMS1880005-supplement-Supplemental_Materials.docx^ (15.1MB, docx) Data file S1 [383]NIHMS1880005-supplement-Data_file_S1.xlsx^ (142.7KB, xlsx) Acknowledgments: