Single-Cell Analysis of Tumor-Specic Neutrophils During Gastric Cancer Progression Inspired a Hippo Targeted Antitumor Therapy

Neutrophils are important for tumor immunosurveillance; yet their roles in tumorigenesis remain debating due to incomplete understanding of their heterogeneity and functional plasticity. Here we resolve at single-cell resolution the heterogeneity of neutrophils in tumor-bearing mice and patients with gastric cancer. Our results define CXCR2 neg CD44 neg neutrophils as a tumor-specific (tsNeu) population with activated (CD14 high ) but atavistic phenotypes (CD53 high CD63 high ). Human and mouse tsNeus can be further classified into four subsets according to their expression of CD54 and CD101. Importantly, the spatiotemporal evolution and anti-tumor function of tsNeu requires the Hippo pathway transcriptional co-activator YAP/TAZ. Deficiency of YAP/TAZ impairs the generation of effective tsNeus and therefore accelerates tumor progression. Remarkably, treatment with the Hippo antagonist phosphatidic acid (PA) turns blood neutrophils into tsNeus with enhanced abilities of direct killing tumor cells, as well as inducing adaptive anti-tumor immunity. Collectively, our work not only identified molecular makers for tsNeus, but also revealed key transcriptional programs for their fate determination and function; and further offers a strategy of targeting Hippo-YAP/TAZ for neutrophil therapy against gastric cancer.


INTRODUCTION
Neutrophils serve as the first line of host defense against pathogen infection and tissue injury.
In response to inflammatory stimuli, neutrophils migrate from the peripheral blood to infected or injured tissues 1 . Neutrophils are also found in the tumor microenvironment and associated with tumor progression in a variety of cancers 2 . However, the specific role of neutrophils in cancer initiation and progression has long been under controversy: tumor associated neutrophils appear to exert dual functions with either anti-tumor or pro-tumor effect.
The anti-tumor effects of neutrophils include not only cytotoxicity and innate immunity, but also the induction of adaptive immunity 3,4 . For example, neutrophils directly kill tumor cells via ROS production and H2O2-mediated cytotoxicity 5,6,7 . As a key player of innate immunity, neutrophils can restrict bacterial infiltration to inhibit intestinal microbiota driven tumor-promoting inflammation and colorectal carcinogenesis 8,9 . Moreover, tumor-associated neutrophils acquire an activated phenotype with characteristics of antigen-presenting cells to stimulate conventional CD4 + and CD8 + T cell proliferation and production of effector cytokines 10,11 . In addition, neutrophils may cooperate with macrophages to promote the activation of CD4 -CD8unconventional αβ T cells and IFNγ production 12,13 , as well as to suppress interleukin-17A (IL-17A) production by tumor-promoting γδT cells 13 .
Neutrophils can also exert pro-tumor effects including angiogenesis, extracellular matrix remodeling, distant metastasis and immune-suppression. For instance, neutrophils promote angiogenesis by the production of pro-angiogenic factors 14,15 . Neutrophil extracellular trap (NET)-associated proteases such as elastase MMP9 cleave the extracellular matrix protein laminin and induce the proliferation of dormant cancer cells 16 . Neutrophils can escort circulating tumor cells (CTCs) and drive cell cycle progression within the bloodstream to enhance the metastatic potential of CTCs 17 . Moreover, DNA component of NETs acts as a chemotactic factor to attract cancer cell for distant metastasis 18,19 . Neutrophils not only inhibit natural killer cytotoxicity to tumor cells 20 , but also suppress effector T cell-mediated antitumor response by production of ROS, NO, prostaglandin E2 and upregulation of programmed cell death 1 ligand 1 (PDL1) 21,22,23 .
The paradoxical phenotypes of neutrophils largely arise from the enormous spatiotemporal heterogeneity concerning their maturation, activation, half-life, aging status, tissue distribution and effector function diversity 24,25 . Neutrophils are derived from common myeloid progenitors (CMPs) through granulocyte monocytes progenitors (GMPs) 26 . During differentiation and maturation, neutrophils undergo a series of morphological changes: myeloblast, promyelocyte, myelocytes, metamyelocytes and band cells 27 . Recent single cell RNA sequencing (scRNA-Seq) analysis defined eight neutrophil subpopulations from bone marrow, peripheral blood and spleen, with distinct maturing bone marrow neutrophil subsets giving rise to different mature peripheral neutrophil subsets 28 . Moreover, a very early unipotent neutrophil progenitor termed ''eNeP'' has been defined as Lin -CD66b + c-Kit + CD71 + in human bone marrow, in parallel with GMPs of mice 29 . Meanwhile, scRNA-Seq analysis identified Ly6C + CD115 -CD81 + CD11b -CD106 -GMP as an early committed neutrophil progenitor (proNeu1) 30 . ProNeu1 sequentially differentiates into proNeu2 and then neutrophil precursor (preNeu), which are committed to non-proliferative immature and then mature neutrophils 31 .
Differentiation and maturation of neutrophils are orchestrated by various transcription factors involved at different stages of development 28 . For example, C/EBPα which is highly expressed in GMPs, not only plays an import role in the initiation of granulopoiesis, but also is required for the differentiation of proNeu1 to proNeu2 and then preNeu 31 . Growth factor independent 1(GFI-1), which is highly expressed in proNeu2 and preNeu, is essential for early neutrophil differentiation 32 . PU.1 and C/EBPδ are involved in terminal granulopoiesis and expressed in bone marrow mature neutrophils, as wells as in peripheral blood circulating neutrophils. Various granule proteins are differentially synthesized and stored: primary granules such as Mpo, Elane and Ctsg are highly expressed in GMPs and proNeus; secondary granules such as Ltf, Camp and Lcn2 are expressed within preNeus and bone marrow immature neutrophils; tertiary granules such as Mmp8, Mmp9 and Itgam are formed within bone marrow mature and peripheral neutrophils.
In contrast to the well-studied neutrophil development under physiological condition or infection, the maturation and activation of neutrophils are poorly understood in the case of tumorigenesis. For example, neutrophil heterogeneity remains to be fully characterized in the tumor microenvironment; the evolution routes during tumor progression and the transcriptional program for their fate decision are largely unknown. Therefore, therapeutic targeting of neutrophils is underdeveloped. Here, we performed a comprehensive study of neutrophils associated with gastric cancer (GC) via combining scRNA-Seq with mice tumor model, and clinical GC sample, which defined tumor-specific neutrophils (tsNeu) as a CXCR2 neg CD44 neg population with four subsets marked by CD54 and CD101. We further demonstrated that YAP/TAZ are required for the generation and antitumor function of tsNeus.
Based on these findings, we developed a Hippo-YAP-targeted neutrophil therapy to treat GC.

scRNA-Seq Analysis of Ly6G + Neutrophils in Mice Bearing GC
To characterize tissue-specific neutrophils during tumorigenesis, we isolated by fluorescenceactivated cell sorting (FACS) CD45 + CD11b + Ly6G + cells to exclude Ly6G neg neutrophiles (GMP and proNeu) from bone marrow (BM), spleen (SP), peripheral blood (PB) and gastric cancer (GC) tissue of mice bearing GC (Fig. 1a). After rigorous quality control, we obtained 22,069 high-quality cells with an average of 1,189 genes per cell profiled, with a total of 17,458 genes detected (Supplementary Fig. 1a-c, Supplementary Table 1). Using genome-wide correlations between cluster mean expression and previously defined transcriptional profiles of sorted neutrophil subsets 30,33,34 and canonical markers (Supplementary Table S2), we determined nine major cell clusters (Fig. 1b upper). Note that both of the two clusters of Ly6c hi Itgb3 hi preneutrophil (preNeu1/2) are mainly from bone marrow and spleen; while both the Ngp hi Cxcr2 lo immature neutrophil (immNeu) and the Cxcr2 hi Mmp8 hi mature neutrophil (mNeu) are mainly from spleen and peripheral blood (Fig. 1B, Supplementary Fig. 1d). These findings are consistent with previous studies 30,33,34 and verify the robustness of the scRNA-Seq analysis in this work.
In contrast to the overlapping of cell clusters from bone marrow, spleen and peripheral blood, the transcriptomic map of most tumor-specific neutrophils (tsNeus) appears to be well separated from that of neutrophils in other tissues (Fig. 1b lower). Dimensional reduction with uniform manifold approximation and projection (UMAP) further revealed four distinct clusters of tsNeus, which we termed tsNeu1-4 ( Fig.1b, Supplementary Fig. 1d). Moreover, we found the clusters of tsNeu1/2 were Cd44 lo Icam1 lo Gdf15 lo , whereas the clusters of tsNeu3/4 were Cd44 lo Icam1 hi Gdf15 hi (Fig. 1c, Supplementary Fig. 1d, Supplementary Table 2). In addition, we also noticed a highly heterogeneous cell cluster consisting of neutrophils from all four examined tissues, which we termed silent neutrophils (sNeu) expressing early markers for neutrophil development such as Elane and Prtn3 (Fig. 1b, Supplementary Fig. 1d,

Supplementary Table 2).
To resolve the lineage of tsNeu relative to neutrophils in other tissues, we performed Monocle pseudo-time analysis 35 . Consistent with the above UMAP analysis, tsNeu was found to be closely associated with neutrophil subsets in peripheral blood (immNeu and mNeu) but more remotely associated with those in bone marrow and spleen (preNeu1/2) (Fig. 1d).
Furthermore, we performed principal-component analysis (PCA) to compare gene expression profiles of all the above identified neutrophil subsets. Consistent with previous studies 33, 34 , the subsets preNeu1, preNeu2, immNeu and mNeu displayed distinct gene signatures (Fig. 1e).
Neutrophil differentiation occurred on a tightly organized trajectory, starting from GMP cells in bone marrow and spleen to mNeu in spleen and peripheral blood. A cluster of immNeu cells followed preNeu expansion and expressed secondary granule genes such as Ngp ( Supplementary Fig. 1d), and this cluster migrated from bone marrow to the peripheral blood.
Neutrophil differentiation in bone marrow tended to end with mNeu highly expressing Mmp8 (Supplementary Fig. 1d). RNA velocity analysis, a method inferring precursor-progeny cell dynamics revealed that preNeu subsets exhibited a strong directional flow towards immNeu and mNeu (Fig. 1f), indicating that neutrophil maturation (from preNeu to mNeu) follows a single main branch without significant division. We also observed a clear directional flow from immNeu and mNeu enriched in peripheral blood to tsNeu enriched in tumor tissue ( Figure 1F).
Neutrophil contains an assembly of granules destined for regulated secretion, each granule type with distinct constituents formed before terminal differentiation 36 . The earliest granules are designated primary (azurophil), followed in time by secondary (specific), and tertiary (gelatinase/secretory) granules 37 . To gain insights into the functional processes of Neutrophil that occur during tumor infiltration, we further analyzed the expression of various granule genes in tsNeus, the major population of neutrophils in the tumor tissue. tsNeus showed a highly coordinated expression of tertiary granules at levels comparable to that of mNeu (Fig. 1g). We also measured the levels of neutrophil maturation, aging and apoptosis based on their expression of related genes (Supplementary Table 3). Overall, the maturation levels of tsNeu are similar to those of immNeu and mNeu (Fig. 1g). Notably, tsNeu3/4 showed the highest scores of aging and apoptosis among all tumor-specific neutrophils, indicating a terminal bound identity (Fig. 1g, Supplementary Fig. 1e). tsNeu4 cluster showed the highest expression of key genes involved in reactive oxygen species (ROS) production, phagocytosis and chemotaxis (Fig. 1h).
Taken together, these results indicate that tsNeus represent a unique neutrophil population terminal to the trajectory along tumorigenesis; and that tsNeus are activated but exhibit an atavistic transcriptomic signature.

Identification and Isolation of tsNeus by CD44 and CXCR2
Our findings of tsNeu at the level of transcriptomic signatures promoted us to further characterize at the protein level the cell identity and state of this population. From the flow cytometric analysis, we observed preNeus (CD11b + Gr1 + CXCR4 + c-Kit + ) in tumor-bearing mice with a gradual upregulation of c-Kit and a subsequent bifurcation of c-Kitneutrophils into CD101and CD101 + cells, representing the previously characterized subsets of immNeu and mNeu, respectively (Supplementary Fig. 2a) 33 . Given the common origin of Ly6G + ly6C -neutrophils, however, these surface markers could not isolate tsNeus from other types of neutrophils ( Supplementary Fig. 2a, b), challenging further study of distinct neutrophil subsets during tumorigenesis. To solve this issue, we systematically evaluated the expression levels of a group of cell surface markers, including neutrophil lineage-specific (c-kit, CD53, CD63, CD101, CXCR2, CXCR4, Ly6G, Ly6C), activation (CD14, CD54), rolling (CD44, CD62L), trans-endothelial migration (CD11b) markers 1,38,39 .

Time-based Evolution Trajectory of tsNeus along Tumor Progression
To investigate the dynamic evolution of tsNeus along tumor progression, we analyzed the transcriptomic network of neutrophils from mice bearing GC for 3 or 10 days. With the same sequencing depth of scRNA-Seq, we observed an increase in both gene number and total UMIs in neutrophils isolated from mice bearing tumors for 10 days when compared with those from mice bearing tumors for 3 days, indicating elevated transcriptional activity in neutrophils during tumorigenesis (Supplementary Table 1). We performed t-distributed stochastic neighbor embedding (t-SNE) analysis 40 and confirmed that tumor-infiltrated neutrophils were a mixed population of preNeu, immNeu, mNeu, tsNeu, and sNeu (Fig. 3a). We observed a decreased percentage of tsNeu1/2 but increased percentage of tsNeu3/4 in mice bearing tumors for 10 days when compared with those for 3 days (Fig. 3a, b), indicating transition of tsNeu1/2 into tsNeu3/4 along tumor progression. Consistent with the results of the above scRNA-Seq analysis, these clusters possessed key signature genes like CD63, CD54 (ICAM1) and Gdf15, and their gene expression correlated with the presence of tsNeu1/2, tsNeu3 and tsNeu4, respectively (Fig. 3c).
Subsequently, we verified at protein levels the expression of various surface markers in tumor-infiltrated neutrophils. We observed a significantly decreased number of total tumorinfiltrated neutrophils (Ly6G + Cd11b + Ly6C lo ) in mice bearing GC for 10 days when compared with that for 3 days (Fig. 3d), indicating a rapid loss of tumor-infiltrated neutrophils along the disease progression. In contrast, we found the ratio of CD44 -CXCR2 -tsNeu to total tumorinfiltrated neutrophils significantly increased from ~88% in mice bearing GC for 3 days to ~97% in mice bearing GC for 10 days (Fig. 3e). In addition, we observed no significant change in the CD115 + Ly6G + cells during tumorigenesis, a previously reported population of MDSCs 41,42 ( Supplementary Fig. 3a). In keeping with the scRNA-Seq results (Fig. 3a, b), our flowcytometry analysis also revealed a similar change of the tsNeu population along tumor progression: decreased number of CD54 -tsNeu (tsNeu1/2) but increased number of CD54 + tsNeu (tsNeu3/4) in mice bearing GC for 10 days than those for 3 days (Fig. 3f).
Notably, the expression of CCL5, CD14, CD53 and CD63 in CD54 + tsNeu (tsNeu3/4) from mice bearing GC for 10 days were significantly higher than that for 3 days (Supplementary Fig.   3b, 3c). In addition, we used early neutrophil-lineage markers, including CD34 and c-Kit, to isolate sNeu from total tumor-infiltrated neutrophils (Ly6G + Ly6C lo ). As shown in Supplementary Fig. 3d, the proportion of sNeu (Ly6G + CD34 + c-Kit + ) was significantly reduced from ~2.0 % in mice bearing GC for 3 days to 0.27% in mice bearing GC for 10 days, findings consistent with our scRNA-Seq analysis. Taken together, these results revealed a stepwise and hierarchical trajectory for tsNeu evolution during tumorigenesis.

YAP/TAZ Is Required for the Antitumor Function of tsNeu
To gain insights into the global gene regulatory networks involved in the lineage commitment of tsNeus, we performed SCENIC (single-cell regulatory network inference and clustering) analysis 43 to compare the regulon activities between tsNeu and other neutrophil subsets, which identified 20 neutrophil-specific networks (Fig. 4a). As anticipated, preNeus highly expressed Gfi1 and Cebpe regulons, which are necessary for the early neutrophil development 32 (Fig. 4a). In contrast, tsNeus highly expressed 6 regulons, including Egr1, Fosb, Balf, Irf5, Atf3 and Jun, which are associated with the activation of neutrophil 34 (Fig. 4a). Notably, clustering analysis revealed high activity of Tead regulon in tsNeus (Fig. 4A). Moreover, the pattern of Tead regulon was similar to that of Atf3 and Jun (Supplementary Fig. 4a). Further gene set enrichment analysis (GSEA) demonstrated a global transcriptome change in tsNeus (relative to preNeus, immNeus, mNeus), with a significant positive enrichment of target genes of the YAP/TAZ-TEAD complex (Fig. 4b). Consistently, immunoblotting analysis showed that tumor-infiltrated Neu have the highest protein levels of YAP and TAZ in the four tissue samples (Fig. 4c). Importantly, the increase of YAP and TAZ reached a peak after implantation with MFC for 10 day, and then decreased until 15 days (Fig. 4d). In addition, tsNeu1/2 subsets (DN, SP1) showed higher activity of YAP/TAZ-TEAD complex when compared to tsNeu3/4 subsets (SP2, DP) (Fig. 4e), findings consistent with the above observations that tsNeu1/2 had lower apoptosis than tsNeu3/4 ( Supplementary Fig. 1e, Fig. 2f).
Next, we cross-bred Yap1 flox/flox mice and Taz flox/flox mice with LyzM Cre mice (termed as Yap1;Taz DKO ) to generate mice with neutrophils deficient of YAP and TAZ (Fig. 4f). We did not observe a significant change for the development and total number of neutrophils in the Yap1;Taz DKO mice compared to those of wildtype control littermates (Supplementary Fig. 4b).
In contrast, in a subsequent xenograft tumor model, we observed much larger tumors in Yap1;Taz DKO mice than in control mice, even though same amount of MFC tumor cells were inoculated (Fig. 4g). To rule out the possibility that such results were caused by the nonspecific effects of LyzM Cre , we adoptively transferred wildtype or Yap1;Taz DKO neutrophils into wiltype mice loaded with tumors (Fig. 4h). As expected, the tumor volume was decreased after 3 times of intratumoral injection of neutrophils derived from wildtype mice when compared to the mock group (Fig. 4i). However, such effect was not observed for the group of neutrophils derived from Yap1;Taz DKO mice, indicating a defect of tumor-killing ability for these neutrophils (Fig. 4i). Together, these results clearly indicate that YAP/TAZ is required for the antitumor function of neutrophils.

YAP/TAZ Is Essential for Lineage Commitment of tsNeu
We then examined the regulatory effects of YAP/TAZ towards various subsets of neutrophils during tumorigenesis. In the late stage of tumor progression, we observed a significantly decreased ratio of total tumor-infiltrated neutrophils (Cd11b + Ly6G + Ly6C lo ) (Fig. 5a,   Supplementary Fig. 5a), but a significant increased ratio of tumor-infiltrated preNeus (ckit + CXCR4 + ) in Yap1;Taz DKO mice (Supplementary Fig. 5b). However, YAP/TAZ deficiency did not affect the ratio of GMPs, preNeu, immNeu and mNeu in the bone marrow of tumorbearing mice (Supplementary Fig. 5c). Notably, the ratio of CD44 -CXCR2 -tsNeu to total tumorinfiltrated neutrophils significantly decreased from ~97% in control mice to ~41% in Yap1;Taz DKO mice after bearing tumors for 20 days (Fig. 5b). Meanwhile, we observed an increased number of CD101 + CD54 -/tsNeu2 but a decreased number of CD101 -CD54 + /tsNeu3 and CD101 + CD54 + /tsNeu4 in Yap1;Taz DKO mice compared to the wildtype control mice after bearing tumors for 10 days (Fig. 5c). Moreover, YAP/TAZ deficiency reduced the expression levels of activation markers CD14, but enhanced the levels of early neutrophil lineage markers CD63 and c-Kit (Fig. 5d).
To investigate Hippo signaling-mediated immune roadmap to tumor progression, we isolated and sequenced a total of 10,858 CD45 + single cells from the tumors (n=2) of MFCinoculated Yap1;Taz DKO mice and control mice (Fig. 5e) (Fig.   5f). We observed a decreased percentage of total neutrophil but increased percentage of other cell types in the Yap1;Taz DKO mice compared to the control mice ( Fig. 5f, g,   Supplementary Fig. 5d). Consistent with the above observations (Fig. 5c, Supplementary Fig.   5b), YAP/TAZ deficiency led to an obvious decreased percentage of tsNeu3/4 but increased percentage of preNeu (Supplementary Fig. 5e, f). Moreover, tumor-infiltrated neutrophils from Yap1;Taz DKO mice displayed a downregulated transcriptional profile of the activation signature genes like CD14, CD54 (ICAM1), JunD and Nfkb1, but an upregulated profile of early lineage signature genes like Hexa, Ly6c2 and Prnt3 (Fig. 5h). Furthermore, KEGG enrichment analysis of Yap1;Taz DKO versus control mice indicated that the major pathway related to development was Hematopoietic cell lineage, and pathways related to immune activation were the Toll-like receptor signaling, T cell receptor signaling and RIG-I-like receptor signaling (Fig. 5i).
Taken together, these results demonstrate that the YAP/TAZ is essential for the development of tumor-infiltrated neutrophils, and that YAP/TAZ-deficient neutrophils are atavistic and defect in activation.

tsNeu Subsets Are Conserved from Mice to Human
To determine whether tsNeus were also present in human, we performed inDrop scRNA-seq analysis of CD45 + cells isolated from clinical samples of GC patient (Fig. 6a). The gastric cancer tissue (GC), peritumoral tissue (PT) and peripheral blood (PB) were from one patient who had not yet received cancer treatment. After filtering scRNA-seq data to exclude putative cell doublets and stressed or dead cells, a total of 20,985 cell transcriptomes were visualized for exploration via UMAP. Of all cells, 16% of them (n = 3,357) were classified as neutrophils, and other cells of them were identified as monocytes (1,176), macrophages (1,859), T cells (6,266), (Fig. 6b). We then focused on neutrophils, which were resolved into six subsets (immNeu, mNeu, and tsNeu1~4) by spectral clustering (Fig. 6c). Notably, the relative proportions of these neutrophil subsets vary considerably in different tissues: immNeu and mNeu were enriched ~3 fold in PB in comparison to PT and GC; tsNeu1/2 was increased >10 fold in PT and GC in comparison to PB; tsNeu3/4 were exclusively present in GC (Fig. 6c).

NK cells (1,506), B cells (3,969) and dendritic cells (894), respectively
To further define how the neutrophil subsets relate to each other across species, we compared their transcriptional signatures between human and mouse (Fig. 1C,   Supplementary Fig. 6d). In either human or mouse, tsNeu1/2 expressed high amounts of neutrophil lineage markers (CD63 and CD81) and type I interferon genes (ISG15 and RSAD2), whereas tsNeu3/4 expressed activation markers (ICAM1 and CCL5) (Fig. 1c, 6d). Consistently, tsNeus also showed a highly coordinated expression of tertiary granules (Supplementary Fig.   6a, b). Regardless of species, the tsNeu3/4 subsets showed higher scores of maturation, aging and apoptosis, compared to tsNeu1/2 (Supplementary Fig. 6c). Moreover, trajectory analysis showed a common ordering of neutrophil subsets from tsNeu1 via tsNeu2, to tsNeu3 and tsNeu4 (Supplementary Fig. 6d). In addition, ligand-receptor interaction analysis 44 , a method to construct a cell-cell affinity matrix revealed high affinity of neutrophils with monocytes in the tumor microenvironment (Fig. 6e).
Taken together, these results indicate that tsNeus are conserved in human and mice, and that the tsNeu subsets were differentially associated with tumor progression.

A Hippo-directed Neutrophil Therapy Inhibits Tumor in Combination with PD-1 Blockage
Given the enhanced tumor-killing ability of CD54 + tsNeus and our findings that YAP/TAZ is essential for the production and function of such tsNeus, we reasoned that pharmacologic activation of YAP/TAZ may reshape neutrophils for cancer immunotherapy. To test this hypothesis, we treated mice with phosphatidic acid (PA), a known Hippo pathway regulator that activates YAP/TAZ via intraperitoneal injections (50, 100mg/kg) (Fig. 7a). After injection of PA for 3 consecutive days, we observed an increase in the number of total neutrophils in the murine blood (Fig. 7b). We then isolated neutrophils from the blood of PA-treated mice, found that PA-treatment dose-dependently promoted their tumor-killing ability (Fig. 7c) and the fraction of CD44 -CXCR2 -Neu in blood (Fig. 7d). Consistently, in a nude mice xenograft tumor model, the growth of MFC-derived tumors was significantly inhibited in mice receiving PA treatment when compared to the control group (Fig. 7e). We also observed an increase in the number of total neutrophils in blood (Fig. 7f), as well as the number of CD54 + neutrophils in tumor tissue (Fig. 7g). Moreover, tumors-bearing mice adoptively transferred with neutrophils from PA-treated mice showed much smaller tumors when compared to the control group (Fig. 7h).
To further evaluate the immunotherapeutic effects of targeting the Hippo pathway, we isolated CD45 + CD11b + Ly6G + neutrophils from peripheral blood of four GC patients at different tumor stages, as well as Epcam + tumor cells from these patients. We then treated the isolated neutrophils with PA for 24 h, and subsequently incubated with tumor cells isolated from the same patient (Supplementary Fig. 7a). We observed that PA treatment strongly increased killing ability of blood neutrophils toward GC of different tumor stage (Fig. 7i). Similar to the observations in mice, PA treatment also significantly increased the percentage of tsNeus (CD44 -CXCR2 -) in human PB neutrophils stimulated with HGC-27, a human GC cell line (Fig. 7j,   Supplementary Fig. 7b). Furthermore, we tested a combined therapy of PA and anti-PD-1 antibody in a mouse xenograft tumor model, in which tumors were pre-implanted for 5 days before the treatment. The combined therapy showed a better efficacy than monotherapies in inhibiting tumor progression (Fig. 7k). Note that PA monotherapy had a marginal effect on CD8 + T cells after CD3/CD28 stimulation (Supplementary Fig. 7c).
Taken together, these results indicate that targeting the Hippo pathway may reprogram neutrophils for cancer immunotherapy, and its combination with anti-PD-1 blockage can further boost the therapeutic effect. Interestingly, our study also identified a population of silent neutrophils (sNeu) expressing early lineage markers (CD34 + c-Kit + ) with potential to differentiate into tsNeus, a phenomenon reminiscent of reserved stem cell. In this regard, a preview report proposed that CD34 + c-Kit + early neutrophils with pro-tumoral activity were accumulated in a murine tumor model 48 . The lineage specification of sNeu, as well as their ability to differentiate into tsNeus remain to be clarified. Given the essential roles of YAP/TAZ in tsNeu development, it is also possible that the Hippo-YAP signaling regulates the potential stemness of sNeus and their differentiation into tsNeus.

DISCUSSION
In summary, our study provides a comprehensive landscape of neutrophils diversity and function in gastric cancer, and identified molecular makers for tumor-specific neutrophils.
Moreover, we identified YAP/TAZ as key transcriptional coactivators required for tsNeu fate determination and function. These findings not only elucidate the functional importance of YAP/TAZ regulation of tsNeu, but also showcase a new type of immunotherapeutic approach via pharmacological targeting the Hippo-YAP pathway in neutrophils.

Cells
MFC cells were obtained from cell line resource of National Infrastructure (Beijing, China) and were grown in RPMI1640 (Invitrogen/Thermo Fisher Scientific, MA, USA), maintained in culture supplemented with 10% heat-inactivated FBS (Biological Industries), and 1% penicillin/streptomycin (Gibco/Thermo) at 37 °C with 5% CO2 in a humidified incubator (Thermo). Cells were passaged for ≤ 3 months from the frozen early-passage stocks that had been received from the indicated sources. During the study, all cell cultures were periodically tested for mycoplasma using MycoAlert™ Mycoplasma Detection Kits (Lonza, ME, USA).

Mice
All mice were housed under specific pathogen-free conditions in automated watered and ventilated cages on a 12-hr light/dark cycle and handled in accordance with were the guidelines of the Institutional Animal Care and Use Committee of the Institute of Biochemistry and Cell Biology. The approval ID for the use of animals was SIBCB-NAF-14-004-S329-023 issued by the Animal Core Facility of SIBCB.
Mice used in this study were from SLAC Laboratory Animal (Shanghai, China): 6-week-old BALB/c mice (male and female); 6-week-old C57BL/6J mice (male and female). Tumors were allowed to establish, sizes (average 80~120 mm 3 ) were matched and then mice were randomly allocated to groups of 6~10 animals. No blinding was used in the treatment schedules for these experiments since the different treatments were identified by mark on tail. Based on our previous experience, groups of 6~10 animals were used to have sufficient animals per group to provide statistically significant data while keeping the number of animals used to a minimum. Tumor size was determined by caliper measurements of tumor length, width and depth and tumor volume was calculated as volume = 0.5236×length×width×depth (mm 3 ).

Murine Tissue Preparation
All murine tissues in this study were prepared from mice sacrificed by cervical dislocation at

Flow Cytometry and Cell Sorting
Antibodies were purchased from BD, Biolegend, eBioscience or Thermo Fisher (Table SXX). using a FACS AriaIII cell sorter and Diva software (both BD Biosciences) to achieve > 98% purity.

Neu Killing Assay
Murine MFC cells were seeded at 3,000 cells/well in 96-well microtiter plates with 100 1616µl culture medium for 24 h and then cocultured with FACS-purified murine Neu (9,000 cells/well) for the following 48 h. An ATP-based CellTilter-LumiTM Plus cell viability assay kit (Beyotime) was used for detecting tumor-killing ability of Neu. 100μl of the reagent was added into each individual well and mixed for 10min at room temperature, then intracellular ATP content was measured using a luminescence detector (GloMax® 20/20, Promega). Killing ability of Neu was calculated using the following equation: % Killing ability = 100 − [value (test) -value (blank)] × [value (control) − value (blank)] -1 × 100.

Murine Strains
Yap1 Floxed ;Taz Floxed mice has been described (XX). Genomic DNA extracted from tail biopsies was used to evaluate offspring genotype. LysM cre mice were purchased from Jackson. LysM cre mice were crossed with Yap1 Floxed ;Taz Floxed mice to generate Yap1 Floxed ;Taz Floxed ;LysM cre/cre . Successful YAP/TAZ deletion in Neu was confirmed by immunoblotting.

Single Cell RNA (sc-RNA) Sequencing
Designated cells were sorted into PBS containing 0.05% BSA following the 10× Genomics protocol. The cell preparation time before loading onto the 10×Chromium controller was< 2 h.
Cell viability and counting were evaluated with trypan blue by microscopy, and samples with viabilities >85% were used for sequencing. Libraries were constructed using the Single Cell 3' Library Kit V2 (10× Genomics). Transcriptome profiles of individual cells were determined by 10× Genomics-based droplet sequencing. Once prepared, indexed complementary DNA (cDNA) libraries were sequenced with paired-end reads on an Illumina NovaSeq 6000 (Illumina). The GEO accession number for the high-throughput sequencing reported in this paper are GSE168537.

Immunoblotting
Cell lysates were resolved in SDS-PAGE gels. The proteins were transferred to a PVDF membrane (Bio-Rad) and further incubated with antibodies.

Statistical Analysis
Both cellular and animal studies have tended to be underpowered. Estimation of sample size for planned comparisons of two independent means using a two-tailed test were undertaken using the SAS statistical software package (9.1.3). Data are presented as mean ± S.D. for continuous variables and as frequencies and proportions for categorical variables. Continuous data were compared using Student's t-tests (comparing two variables) or one-way ANOVA analysis (comparing multiple variables). Two to three biological replicates were used throughout the study.