Graphical abstract

   graphic file with name fx1.jpg
   [37]Open in a new tab

Highlights

     * •
       Immunosuppression in the tumor microenvironment of O and P
     * •
       A 30-gene signature of ovarian-derived TAMs was validated to
       predict O
     * •
       Dynamic changes in immune and stromal cell states during the
       PT-to-M progression
     * •
       Identification of new transcription factor CLOCK in the TGPRNs of
       metastatic GC
     __________________________________________________________________

   Biological sciences; Cancer; Transcriptomics

Introduction

   Gastric cancer (GC) is a major global malignancy, ranking fifth in
   cancer incidence and fourth in cancer-related deaths, with particularly
   high prevalence in Asia, Eastern Europe, and Central America, where
   incidence rates are notably elevated.[38]^1 Despite the implementation
   of surgical interventions, the prognosis of GC remains poor, with an
   overall five-year survival rate of less than 30%, primarily due to
   late-stage diagnoses and challenges associated with organ-specific
   metastasis.[39]^2

   Metastasis is a major contributor to the poor prognosis of GC, yet the
   molecular mechanisms underlying this cellular behavior remain largely
   elusive.[40]^3 Tumor cells possess the ability to invade distant sites
   via various routes, including blood circulation, the lymphatic system,
   direct infiltration, and transcoelomic spread.[41]^4 During this
   metastatic process, tumor cells undergo adaptive changes, activating
   specific genes and pathways that enable them to thrive in particular
   organs. The interactions between the primary tumor, the host
   microenvironment, and immune cells recruited by the tumor play a
   pivotal role in driving metastatic progression.[42]^5 To develop
   targeted therapeutic strategies and identify potential biomarkers for
   clinical diagnosis, it is crucial to accurately delineate the
   organ-specific features of metastasis. However, the current
   understanding of organ-specific metastasis in GC remains incomplete.

   The exploration of GC diversity has advanced significantly with the
   application of single-cell transcriptome sequencing, a cutting-edge
   methodology.[43]^6 While the conventional approach to investigating
   organ-specific metastasis typically involves transcriptome profiling,
   some researchers have employed bulk RNA sequencing (RNA-seq) to
   identify genes critical to site-specific metastases. The expression of
   these genes in the primary tumor has been linked to the recurrence of
   GC metastasis.[44]^7 However, bulk methods have inherent limitations,
   particularly in discerning cell diversity, resulting in an oversight of
   the intricate intra- and inter-tumor complexity in GC.[45]^8 The
   emergence of single-cell RNA sequencing (scRNA-seq) technology has
   revolutionized this landscape, enabling precise and comprehensive
   examination of intra- and inter-tumor heterogeneity across diverse
   cancer types.[46]^9

   Transcription factors (TFs) play a crucial role in regulating gene
   expression and thereby exert indirect control over various cellular
   processes and states. Overexpression of specific TFs has the potential
   to induce significant alterations in cell fate, making them a promising
   approach to cancer diagnosis, prognosis, and treatment.[47]^10
   Therefore, to enhance our understanding of GC and develop potential
   therapeutic interventions, it is essential to identify cell-specific
   TFs in both primary and metastatic GC.

   In this study, we characterized a total of 42,968 high-quality cells
   from six patients across 10 samples to identify seven major cell types
   and 27 unique immune and stromal cell subsets within the TME of primary
   and metastatic GC. Our analysis comprehensively explored cellular
   compositions, developmental trajectories, cell interactions, and
   transcriptional regulation between primary tumors and metastases. These
   findings provide valuable insights into the cellular, molecular, and
   functional changes during GC progression. Furthermore, we constructed a
   cell-specific TFs regulatory network in both primary and metastatic GC,
   enhancing our understanding of the transcriptional mechanisms driving
   the progression from primary to metastatic GC. Our study paves the way
   for future investigations and offers a valuable resource for subsequent
   academic research.

Results

Remodeling the single-cell transcriptional landscape reveals insights into
primary-to-metastatic GC progression

   We performed an in-depth analysis of single-cell transcriptomic data
   obtained from the Gene Expression Omnibus (GEO) dataset [48]GSE163558
   ,[49]^8 which comprises 10 human tissue samples from six patients. Our
   sample cohort included three primary tumor samples (PT1, PT2, and PT3),
   one adjacent non-tumor sample (NT1), and six unique metastatic samples
   (M): liver metastasis (Li1 and Li2), lymph node metastasis (LN1 and
   LN2), peritoneal metastasis (P1), and ovarian metastasis (O1).
   Specifically, PT1 and Li1 were from patient 1 (p1); PT2 and NT1 were
   from patient 2 (p2); O1 was from patient 3 (p3); PT3, Li2, and LN1 were
   from patient 4 (p4); LN2 was from patient 5 (p5); and P1 was from
   patient 6 (p6). After rigorous quality control, 42,968 cells were
   retained for subsequent analysis, with an average of 1,639 genes, 5,529
   unique molecular identifiers (UMIs), and only about 6% mitochondrial
   genes per cell detected ([50]Figures S1B–S1D). Interestingly, the
   proportion of ribosomal genes, as well as their correlation with
   nCount_RNA and nFeature_RNA, was found to be very low. Unsupervised
   clustering, based on transcriptome similarity, identified 12 distinct
   cell clusters, which were visualized using the t-distributed stochastic
   neighbor embedding (tSNE) analysis. The spatial distribution of cells
   from each patient and sample was also obtained ([51]Figure S1A).
   Correlation analysis validated the reliability of the clusters,
   demonstrating that clusters from the same cell lineage exhibited higher
   similarity than those from different lineages ([52]Figure S1E).

   Immune cells accounted for the majority of our analyzed cells from the
   GC samples, particularly in the lymph node and liver, which has also
   been observed in other scRNA-seq studies of cancer.[53]^11 We
   identified seven major cell types using the following markers: T cells
   (e.g., CD3D, CD3E, CD2), B cells (e.g., CD79A, IGHG1, IGHG3), myeloid
   cells (e.g., CD68, CXCL8), natural killer (NK) cells (e.g., KLRD1,
   GNLY, KLRF1), epithelial cells (e.g., EPCAM, KRT19, ELF3), stromal
   cells (e.g., VWF, COL1A1, COL12A1), and hematopoietic stem cells (HSCs)
   (e.g., AREG, ABCC1, GATA2). Intriguingly, gene ontology (GO) analysis
   showed that multiple mast cell-associated pathways were enriched in
   HSCs ([54]Figure S1G), indicating that HSCs have the potential to
   differentiate into mast cells under certain conditions, may be involved
   in mast cell-related immune responses or tumor progression. These cell
   clusters were characterized by marker gene expression and distinct
   physiological functions, exhibited distinct distributions during GC
   primary to metastatic (PT-to-M) progression ([55]Figures 1A–1C and
   [56]S1F and [57]Table S1). Notably, myeloid, epithelial, and stromal
   cells showed decreasing trends during PT-to-M progression, whereas B
   cells exhibited an increasing trend during GC lymph node metastasis
   progression, and T cells showed an increasing trend during both GC
   lymph node and liver metastasis progression ([58]Figures 1A and 1B).

Figure 1.

   [59]Figure 1
   [60]Open in a new tab

   Single-cell atlas of primary tumor, metastasis, and adjacent non-tumor
   samples

   (A) tSNE visualization highlighting seven distinct cell types among the
   42,968 cells (left), delineating the TME cell clusters across different
   tissue groups (right).

   (B) Proportion of each cell type in various samples (NT, PT, M, Li, LN,
   P, and O) and patients (p1-p6).

   (C) Left: heatmap showing the expression profiles of the top five genes
   ranked by LogFC of each cell type. Right: enriched KEGG pathways for
   the marker genes of each cell type.

   (D) Venn diagram for the overlap analysis of the marker genes across
   the seven cell types and five cancer samples.

   (E) Ring heatmap visualizing the marker genes specific to each of five
   cancer samples.

   (F) Dot plot showing the enrichment pathways in the five cancer samples
   based on the DEGs using KEGG enrichment analysis.

   (G) Functional changes during PT-to-M progression.

   Next, we explored the dynamic changes in the TME from additional
   perspectives. At the molecular level, we identified 114 overlapping
   genes as potential markers during PT-to-M progression, revealing a
   correlation between changes in the cellular compositions and the
   expression of cellular marker genes ([61]Figure 1D). CD79A, a marker
   gene of B cells, was specifically expressed in LN, suggesting a link
   between its expression and the increasing trend of B cells in LN
   ([62]Figure 1E). Additionally, we performed Kyoto Encyclopedia of Genes
   and Genomes (KEGG) enrichment analysis on the differentially expressed
   genes (DEGs) in our cancer samples, which provided insights into the
   functional differences between PT and M ([63]Figure 1F and
   [64]Table S2). The activation of the NF-κB signaling, IL-17 signaling,
   and TNF signaling pathways in PT suggested dynamic interactions between
   cancer cells and the immune and inflammatory components of the
   TME.[65]^12 Furthermore, Li exhibited enrichment for the MAPK signaling
   pathway, and O showed specific enrichment for cholesterol metabolism.
   We also observed the activation of ferroptosis in P, a phenomenon
   reported in other studies,[66]^13 indicating the involvement of cell
   death regulation in GC peritoneal metastasis.[67]^14 Together, the KEGG
   pathway enrichment analysis indicated that cells undergo significant
   functional reprogramming during PT-to-M progression, with upregulated
   antigen processing and presentation pathways observed during the
   progression of GC liver and peritoneal metastasis ([68]Figures 1G and
   [69]S1H).

Immunosuppression in the TME of GC ovarian and peritoneal metastasis and a 30
gene signature of ovarian-derived TAMs was validated to predict ovarian
metastasis

   Sub-clustering analyses of immune and stromal cells identified 27 cell
   subsets ([70]Figures 2A and [71]S2A). T cells were divided into nine
   clusters: CD8^+ T cells, naive CD8^+ T cells, naive CD4^+ T cells,
   CD8^+ T effector memory (Tem) cells, natural killer T (NKT) cells,
   regulatory T cells (Tregs), CD4^+ T effector memory (Tem) cells,
   exhausted CD8^+ T cells, and natural regulatory T cells. B cells were
   primarily composed of six subgroups: T cell-like B cells,[72]^8 naive B
   cells, memory B cells, plasma cells, germinal center (GC) B cells, and
   myeloid-like B cells (M-B cells). Myeloid cells were further divided
   into five subsets: neutrophils, monocytes, macrophages,
   tumor-associated macrophages (TAMs), and plasmacytoid dendritic cells
   (pDCs). NK cells were subdivided into four subclusters: NK cell
   progenitors (Pro-NK cells), CD3/CD28-stimulated NK cells, IL7R + NK
   cells, and SLFN13+ NK cells. Stromal cells were categorized into
   endothelial cells, pericytes, and cancer-associated fibroblasts (CAFs).
   Notably, each cluster exhibited a distinct gene expression pattern
   ([73]Figure 2B and [74]Table S3). Correlation analysis revealed that
   clusters from the same cell lineage showed higher similarity to one
   another than to those from other lineages ([75]Figure S2B), confirming
   the reliability of the clustering. The distribution of UMIs in each
   cell lineage is shown in [76]Figure S2C. These results highlight a high
   degree of cell state diversity within the TME of primary and metastatic
   GC.

Figure 2.

   [77]Figure 2
   [78]Open in a new tab

   Subclusters of immune and stromal cells characterized during GC
   progression

   (A) tSNE showing 27 unique immune and stromal cell subsets (upper
   panels) and their corresponding TME cell clusters (lower panels) across
   different tissue groups.

   (B) Heatmap of the expression of subset-specific markers across cell
   subsets.

   (C) Reproducible cell subset distributions across the five cancer
   samples.

   (D) Heatmap showing tissue preferences of clusters in each immune and