Spatial omics study reveals molecular-cellular dynamics of tumor ecosystem in esophageal squamous-cell carcinoma initiation and progression
Zhao Liu, Wenhao Zhou, Lei Li, Congcong Song, Meng Yue, Huilai Lv, Zhenhua Li, Minghao Zhang, Na Li, Jiaqian Wang, Lianmei Zhao, Haitao Luo, Ziqiang Tian
TL;DR
This study uses spatial omics to track how tumors and their surroundings change during esophageal cancer development, revealing new insights into disease progression and potential treatment targets.
Contribution
The study identifies novel molecular and cellular dynamics in esophageal squamous-cell carcinoma using spatial profiling and validates the role of O-GlcNAc transferase in tumor progression.
Findings
Early-stage ESCC involves dysregulation of epidermal development and immune cell shifts.
Late-stage ESCC features sustained PI3K/AKT pathway activation and disrupted tertiary lymphoid structures.
O-GlcNAc transferase is upregulated in late stages, correlates with poor prognosis, and promotes tumor invasion.
Abstract
Deciphering the molecular and cellular dynamics during esophageal squamous-cell carcinoma (ESCC) evolution is critical for elucidating the underlying disease mechanisms and devising rational targeted therapeutic strategies. Here, we perform digital spatial profiling on 32 tissue samples from 18 patients across different ESCC stages. At ESCC initiation, tumor cells undergo coordinated regulation of epidermal development and keratinocyte differentiation, accompanied by increased B cells and decreased T cells. In late stages, the phosphatidylinositol 3-kinase-protein kinase B (PI3K/AKT) pathway is continuously upregulated, and tertiary lymphoid structures are dysregulated. We further verify these findings by multiplex immunofluorescence. Notably, the O-GlcNAc transferase gene, activated exclusively in late stages, correlates with poor prognosis, and its knockdown inhibits ESCC cell…
Genes, proteins, chemicals, diseases, species, mutations and cell lines named across the full text — each resolved to its canonical identifier and authoritative record.
Click any figure to enlarge with its caption.
Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
Figure 6
Figure 7Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsEsophageal Cancer Research and Treatment · Microbial metabolism and enzyme function · Cancer, Hypoxia, and Metabolism
Introduction
Esophageal cancer is one of the most aggressive and lethal malignant tumors worldwide.1 Esophageal squamous-cell carcinoma (ESCC) is the predominant subtype in East Asia,2 and the 5-year survival rate is only 15%–25%.3^,^4^,^5 Due to its rapid progression and poor prognosis, most studies have focused on the molecular indicators in the initiation stage of ESCC.6^,^7 However, the underlying cellular and molecular mechanisms related to the entire evolutionary process from normal tissue to esophageal squamous precancerous lesions (ESPL) and ultimately to tumor and metastasis8 remain largely unclear.
The evolutionary advantage of tumors is influenced not only by tumor cells but also by the tumor microenvironment (TME).9^,^10^,^11 Previous studies on ESCC occurrence and development have predominantly focused on the tumor cells, often neglecting the TME.12^,^13^,^14 Consequently, a thorough investigation of the TME at distinct stages of ESCC is crucial. Liu et al. revealed that cancer-associated fibroblasts within the TME interact with epithelial cells to drive ESCC progression.9 Additionally, changes in the abundance of immune cells significantly contribute to ESCC occurrence and development.10^,^15 However, the correlation between tertiary lymphoid structures (TLSs) and the initiation and development of ESCC is unclear.
In recent years, techniques such as bulk and single-cell RNA sequencing (scRNA-seq) have been widely applied to investigate ESCC.16^,^17^,^18 However, these techniques have inherent limitations, as they fail to preserve spatial information and context, thereby hindering deeper insights into in situ cell communications and spatial ecology.19 The spatial technology, GeoMx digital spatial profiling (DSP), enables the simultaneous detection of gene/protein expression and cellular location within tissues.20 Two studies used DSP to explore the role of epithelial cells and TME in the occurrence of ESCC9^,^14 but did not investigate the molecular and cellular alterations during the development and metastatic stages of ESCC.
In this study, DSP was performed on samples collected from patients in various stages of ESCC, including normal epithelia (NE), ESPL, non-metastasis ESCC (non-mESCC), metastasis ESCC (mESCC), and metastatic lymph nodes (mLNs). During ESCC onset, we identified diverse expression patterns and persistently dysregulated genes and pathways in the epithelial compartment, as well as changes in B and T cell abundance within the TME compartments. We found that key pathways contributed to tumor advancement and validated the function of the O-GlcNAc transferase (OGT) gene in ESCC progression. We further determined that non-epithelial cell interactions within the TME compartments play a significant role in ESCC development. Overall, these findings provide valuable insights into the comprehensive changes in the tumor ecosystem during ESCC initiation and progression, offering additional perspectives for developing therapeutic strategies.
Results
DSP reveals the heterogeneity of ESCC initiation and progression
We prospectively collected 32 tissue samples from 18 patients at various stages of ESCC initiation and progression, including NE, ESPL, non-mESCC, mESCC, and mLN tissues (Table 1). To systematically characterize the molecular alterations of epithelial cells and immune microenvironment during the evolution of ESCC, we used the spatial whole transcriptome atlas (WTA) panel based on DSP technology. As depicted in Figure 1A, DSP includes probe hybridization with immunofluorescent antibodies, region of interest (ROI) selection and segmentation, and whole-transcriptome quantification by sequencing. To specifically study the molecular profiles of the components, we focused on the epithelium and stroma and segmented epithelial cell and immune cell compartments within each ROI. In addition, ROIs with lymphoid structures (LSs; which consist of TLSs in tumor tissues and lymph follicles in mLNs) were also selected based on hematoxylin and eosin (H&E) staining. At each ESCC stage, five types of areas of interest (AOIs) were identified and separately quantified using whole-transcriptome sequencing, including epithelial-cell-enriched (EP, PanCK-positive), macrophage-cell-enriched (MC, CD68-positive), neutrophil-cell-enriched (NC, CD66b-positive), stroma-enriched (ST, triple-negative), and LS AOIs. After rigorous technical signal and background quality control, a total of 165 AOIs were retained for subsequent analyses (Tables S1 and S2).Table 1. Clinical information and sample collection of the patient cohortPatientsTNMGenderAgeSamplesP01T0N0M0male68NE, ESPLP02T0N0M0female78NE, ESPLP03T0N0M0female76NE, ESPLP04T0N0M0male69NE, ESPLP05T0N0M0female65ESPLP06T0N0M0female58ESPLP07T1bN0M0female62non-mESCCP08T1bN0M0female60NE, non-mESCCP09T1aN0M0female72non-mESCCP10T1bN0M0female67non-mESCCP11T1bN0M0female54non-mESCCP12T1aN0M0male76NE, non-mESCCP13T3N2M0male69mESCC, mLNP14T3N1M0female68mESCC × 2, mLNP15T3N2M0female71mESCC, mLNP16T3N1M0male70mESCC, mLNP17T3N3M0male60mESCC, mLN × 2P18T3N1M0male59mESCC, mLNNE, normal epithelia; ESPL, esophageal squamous precancerous lesion; non-mESCC, non-metastasis esophageal squamous cell carcinoma; mESCC, metastasis esophageal squamous cell carcinoma; mLN, metastatic lymph node.Figure 1. Spatial whole-transcriptome profiling of ESPL and ESCC samples(A) Workflow of DSP experimental process.(B) The UMAP plot displaying the clustering results of AOIs from different compartments at multiple stages of ESCC.(C) The heatmap displaying the expression levels of AOI characteristic markers RNA in different compartments.(D) The boxplots showing the expression levels of marker genes in five different compartments at different stages of ESCC. Box plot shows the median and interquartile range, whiskers extend to 1.5 × IQR. DSP, digital spatial profiling; UMAP, uniform manifold approximation and projection; AOI, area of interest; EP, epithelial-cell-enriched; MC, macrophage-cell-enriched; NC, neutrophil-cell-enriched; ST, stroma-enriched; LS, lymphoid structure; CAFs, cancer-associated fibroblasts; ESPL, esophageal squamous precancerous lesion; non-mESCC, non-metastasis esophageal squamous cell carcinoma; mESCC, metastasis esophageal squamous cell carcinoma; mLN, lymph node metastasis tissues.
Dimension reduction based on uniform manifold approximation and projection (UMAP) was performed by analyzing all the spatially defined AOIs (Figure 1B). As expected, clear separation between EP and stromal microenvironment compartments was observed. Further subclustering analyses of EP AOIs using UMAP revealed a close association between the profiled compartments and the histological features of ESPL and ESCC progression stages. Furthermore, we observed that non-mESCC EP compartments were positioned between those of ESPL and mESCC or mLNs. Meanwhile, mESCC and mLNs were clustered together, indicating that they share a similar spectrum of transcriptional programs. In contrast, no clear distinction was observed among the different segmented compartments derived from the stromal microenvironment compartment, suggesting that stromal cells may play a more complex regulatory role in ESPL and ESCC development. Then, we further assessed gene expression programs in the five types of AOIs based on the canonical markers (Figure 1C). Consistent with our segmentation strategy, gene expression in the five segmented tissue areas substantially corresponded to their representative cell types. Accordingly, the high expression patterns of PanCK, macrophage, neutrophil, cancer-associated fibroblast, and B cell gene programs were observed in EP, MC, NC, ST, and LS compartments, respectively (Figure 1D). Furthermore, we employed xCell to quantify the enrichment scores of epithelial cells, stromal cells, macrophages, and neutrophils across different compartments. The enrichment score of the primary cell type for each compartment was consistently aligned with the compartment’s predefined identity (Figure S1A). Additionally, we performed deconvolution analysis on the non-epithelial regions of our DSP data using CIBERSORT and leveraged the scRNA-seq-derived LM22 signature matrix to quantify the proportions of macrophages and neutrophils. The results showed that the proportion of macrophages was significantly higher in macrophage-enriched compartments compared to other compartments, while the proportion of neutrophils was significantly elevated in neutrophil-enriched compartments (Figure S1B). These results show that the areas derived from different tissues were accurately segmented and quantified based on DSP technology, which enable us to systematically and accurately analyze the molecular changes during ESCC initiation and progression in subsequent studies.
Activation of tumorigenesis pathways in EP compartment is associated with the stepwise initiation of ESCC as revealed by spatial transcriptome expression patterns
To gain insight into the mechanisms that promote stepwise tumor initiation, we characterized diverse cellular components in the tumor ecosystems across ESPL and ESCC progression stages considering several aspects, such as their transcriptional patterns, regulatory functions, cellular gene signatures, as well as cellular compositions. We first investigated the altered genes in EP cells from NE to ESPL and found that several ESCC-specific tumorigenesis pathways including keratinocyte differentiation and epidermis development were dysregulated (Figures 2A and 2B; Table S3). Specifically, expression programs such as keratinocyte differentiation and hypoxia-inducible factor-1α signaling pathway showed a significant decrease in ESPL compared to their expression in NE (Figure 2B). Additionally, the expression of keratinocyte-differentiation-related genes such as small proline rich protein 3 (SPRR3), cystatin A (CSTA), and transglutaminase 1 (TGM1) was considerably decreased in the progression from NE to ESPL, consistent with the results of previous studies that showed that aberrant keratinocyte differentiation was a key factor in the initiation of cutaneous and esophageal squamous carcinoma.21^,^22^,^23 By comparison, significantly increased expression of the epidermis development program was observed in ESPL, along with the upregulation of several embryonic-like genes such as SRY-related HMG-box (SOX4) and keratinocyte-differentiation-associated protein (KRTDAP) (Figure 2C). These expression patterns indicate the critical role of coordinated regulation of epidermis development and keratinocyte differentiation in maintaining the normality of esophageal squamous epithelium and promoting the occurrence of precancerous lesions. Combining previous findings and evidence of interactions among proteins associated with epidermis development and keratinocyte differentiation, we deduced that dedifferentiation of keratinocytes may be a key process in the development of ESPL, which is orchestrated by multiple regulatory programs. Furthermore, the activation of embryonic-like genes may be an initial step of keratinocyte dedifferentiation, which further induces epidermis-development-associated genes (such as KRTDAP, calmodulin like 5 [CALML5], keratin 10 [KRT10], and psoriasin [S100A7]), leading to the downregulation of genes involved in keratinocyte differentiation (such as filaggrin [FLG], small proline-rich protein 2F [SPRR2F], annexin A1 [ANXA1], and transglutaminase 3 [TGM3]) (Figure 2D). Notably, recent experiential studies suggest that increased Sox9 and decreased ANXA1 expression are crucial for esophageal cancer initiation, further confirming our findings.16^,^24^,^25Figure 2. Changes in gene expression patterns in the EP compartment during the initiation of ESCC(A) Volcano plot showed significant DEGs during the formation of ESPL.(B) Pathway enrichment analysis of DEGs in ESPL and normal stages.(C) Changes in gene expression across epidermis development and keratinocyte differentiation pathways during ESPL formation.(D) Interaction plot of ESPL-formation-related genes. Each circle represents a protein, and the interactions are connected by solid lines. The types of interactions were shown on the right side of the figure.(E) The four-quadrant plot illustrates the significant differences in gene expression patterns among four distinct types during the initiation of ESCC.(F) The boxplot shows the expression levels of the presented genes across the four patterns during ESCC initiation. Pattern 1 to 4 were tagged in blue, pink, yellow, and purple, respectively. Box plot shows the interquartile, whiskers extend to1.5 × IQR.(G) Pathway enrichment results for genes in the four patterns.(H) The line graph depicts changes in the signature scores of the indicated pathways during the initiation and development of ESCC. DEGs, differentially expressed genes.
We next explored gene programs that may drive the evolution of non-mESCC from ESPL. By simultaneously analyzing the three stages of dysregulated genes, we divided the differential genes into four expression patterns (Figures 2E and 2F; Table S3). There were 36 genes in pattern 1, which were less expressed in NE and gradually increasingly expressed in ESPL and non-mESCC. As presented in the gene set enrichment analysis (Figure 2G), these progressively dysregulated genes were significantly enriched in extracellular matrix (ECM)-receptor interaction, PI3K/AKT signaling pathway, and ECM organization processes, suggesting that the continuous activation of genes such as keratin 17 (KRT17) and secreted phosphoprotein 1 (SPP1) may promote the development of ESCC. Additionally, based on the functional enrichment of genes in pattern 2, the keratinocyte differentiation process consistently declined during ESCC initiation. In pattern 3, genes involved in innate immune response and positive regulation of B cell activation were highly expressed in ESPL but were less expressed in non-mESCC. This suggests that esophageal epithelium cells in ESPL could activate innate and B-cell-mediated immune responses by expressing specific genes such as immunoglobulin heavy constant alpha 1 (IGHA1) and IGHG3, but these genes were then downregulated in the non-mESCC stage. Finally, we investigated the genes in pattern 4 that were downregulated in ESPL but activated in non-mESCC. These genes were found to be involved in the regulation of angiogenesis and hemidesmosome assembly. Therefore, we concluded that dysregulation of keratinocyte differentiation and epidermis development were crucial for tumor initiation, and dedifferentiation of keratinocytes, triggered by embryonic-like genes, played a pivotal role in ESPL development. Additionally, continuous activation of ECM-receptor interaction and PI3K/AKT signaling pathways, along with dynamic immune response changes, drove the progression from ESPL to non-mESCC.
To further dissect whether spatially adjacent epithelial and tumor microenvironment AOIs exhibit functional associations, we performed targeted spatial colocalization analyses. Specifically, we screened AOIs where EP compartments and their adjacent immune compartments co-localized within the same spatial regions, with a focus on the initiation stage of ESCC. Subsequent correlation analyses between ESCC-associated signature genes from epithelial AOIs and immune-regulation-related genes in neighboring immune compartments uncovered stage-specific associations between epithelial gene expression patterns and immunosuppressive features (Figures S2A–S2C). Notably, SPP1, a key mediator of ECM-receptor interaction and PI3K/AKT signaling, showed a notable positive correlation with CD274, a marker of immunosuppressive macrophages, in adjacent MC compartments. This correlation was further strengthened in non-mESCC compared to the ESPL stage (Figure S2D), which indicates that there is a potential spatially coordinated immunosuppressive crosstalk between the expression of SPP1 in epithelial cells and PD-L1-positive macrophages during the initiation of ESCC.
Next, we performed a systematic analysis of the dynamic changes in pathway enrichment results corresponding to different gene expression patterns identified in the epithelial compartment across various stages of ESCC initiation and progression. To quantitatively assess the activity of pathway dynamics, we defined a “signature score” for each pathway, which was calculated as the median expression of all genes enriched in that specific pathway. Our results revealed distinct trajectory patterns of key pathways across ESCC development (Figure 2H). Pathways associated with tumorigenesis and cell proliferation, such as the PI3K/AKT signaling pathway and cell-proliferation-related pathways, exhibited a continuous upward trend in their signature scores from the normal tissue stage to advanced ESCC. This progressive activation was consistent with the gradual upregulation of pattern 1 genes (e.g., SPP1 and KRT17), key molecules that drive the activation of these pro-tumor pathways. In contrast, pathways linked to epithelial homeostasis, including the keratinocyte differentiation pathway (a core pathway of pattern 2) and cell adhesion pathways, showed a steady decline in their signature scores along the same disease progression trajectory. This sustained suppression reflects the loss of normal epithelial cell identity during malignant transformation. The hemidesmosome assembly pathway displayed a unique biphasic dynamic: its signature score decreased from normal epithelium to ESPL but rebounded in advanced ESCC. Collectively, these findings demonstrate the existence of functional trade-offs between the suppression of epithelial differentiation programs and the activation of pro-tumorigenic signaling during ESCC development, further confirming the coordinated molecular reprogramming that drives malignant progression.
Spatial whole-transcriptome analysis deciphers the changes in the stroma microenvironment during ESCC initiation
As tumor initiation and development are strongly influenced by the TME, we further studied the immune infiltration as well as the composition and status of immune cells to explore the changes in the TME during ESCC initiation. We first identified and quantified the immune cells based on spatial WTA data of the PanCK-negative compartment. Statistical analysis revealed that the proportions of B cells and plasma cells were significantly increased in ESPL compared to their proportions in NE tissue (Figures 3A, 3B, and S3A; Table S4), which was further verified by independent analysis (Figure 3C). Notably, we found increased levels of both naive and memory B cells in ESPL (Figures 3C and S3B; Table S4). These observations are consistent with the abovementioned findings on EP compartment in our study and previous studies, in which innate and B-cell-mediated immune responses were activated in the ESPL stage.9^,^16 In addition, the proportions of CD4^+^ and CD8^+^ T cells showed a significant decline in non-mESCC compared to the proportions in the ESPL stage, suggesting that decreased lymphocyte infiltration may promote ESCC initiation (Figures 3A–3C). To further substantiate these findings, we conducted multiplex immunofluorescence (mIF) staining on these samples from normal to non-mESCC stages (Figure S4A); the percentage of CD20-positive cells was significantly increased from the NE to ESPL stages, and the percentage of CD8^+^ and CD4^+^ T cells was notably decreased from the ESPL to non-mESCC stages (Figures 3D and 3E).Figure 3. Spatial transcriptome analysis of the TME during ESCC initiation process(A) Quantitative analysis of immune cells in non-EP compartments during the initiation process of ESCC using TME_consense algorithm.(B) Comparison of B, plasma, CD4^+^ T, and CD8^+^ T cells during the initiation process of ESCC.(C) Analyze the immune cell changes in non-EP compartments during the initiation process of ESCC using SpatialDecon algorithm.(D) mIF staining of CD20, CD8, and CD4 on tumor tissues in ESCC initiation stages. Scale bar, 200 μm.(E) Comparison of the percentages of CD20-, CD8-, and CD4-positive cells across the ESCC initiation stages.(F) Pathway enrichment analysis of significantly differential expression genes in non-EP compartments during the initial stage of ESCC.(G) The heatmap shows the differential genes in the MC and NC compartments during the initial stage of ESCC.(H) Comparison of the proportion of tumor-associated macrophage and neutrophile cells across the ESCC development.(I) The heatmap shows the differential genes in TLS compartments during the initial stage of ESCC.(J) mIF staining on TLS compartment of ESCC initiation stage. Scale bar, 100 μm.(K) Comparison of the percentages of APOBEC3A-positive cells between ESPL and non-mESCC stages in the TLS. NE, normal epithelia; ESCC, esophageal squamous cell carcinoma; mIF, multiplex immunofluorescence; MC, macrophage-cell-enriched compartment; NC, neutrophil-cell-enriched compartment; TLS, tertiary lymphoid structures.Box plots show the median and interquartile range, whiskers extend to 1.5×IQR. ∗p < 0.05*,* ∗∗p < 0.01,∗∗∗p < 0.001, p values were calculated using a two-sided Wilcoxon rank-sum test.
The spatial segmentation strategy we adopted in this study enabled us to further analyze the molecular changes in the MC, NC, and TLS compartments. Although no significant change in the proportion of MC was observed during ESCC initiation process, in-depth analysis of their gene expression patterns revealed that multiple pathways were associated with tumorigenesis (Figure 3F). For example, downregulation of complement activation and immune response and upregulation of cell growth and proliferation were observed in MC during progression from the ESPL to non-mESCC stages, with multiple genes involved in the regulation of these processes (Figures 3F and 3G). Furthermore, NC showed a decreasing trend in ESPL but increased in ESCC (Figure 3C). Genes involved in regulating cytokines and apoptotic process were mainly dysregulated in NC, suggesting their regulatory roles in ESCC tumorigenesis (Figures 3F and 3G). We next conducted subtype-specific analysis of macrophages and neutrophils using a deconvolution approach to investigate the changes in these cell subtypes during the initiation and progression of ESCC. Our results showed that the proportion of pro-tumor tumor-associated neutrophil 2 (TN2) significantly increased from the ESPL stage to the non-mESCC stage, while the proportion of tumor-associated macrophage 2 (TAM2) was markedly elevated in the mLN stage. These subtype-specific changes further confirm that macrophages and neutrophils play stage-dependent functional roles in the initiation and metastasis of ESCC (Figure 3H). Finally, as an important element of the TME, TLSs are ectopic lymphoid aggregates mainly composed of B cells, which play critical roles in tumor adaptive immunity. By analyzing the changes in the expression patterns and regulatory pathways of TLSs, we found that several genes associated with B cells and immune response pathways were dysregulated in TLSs in the non-mESCC stage, and apolipoprotein B mRNA editing enzyme catalytic subunit 3A (APOBEC3A) gene expression showed a marked difference (Figures 3F and 3I). To further validate these findings, mIF was performed using independent samples. Comparison of the protein signal from the TLS compartment in ESPL and non-mESCC showed significantly increased expression of APOBEC3A as the stages progressed (Figures 3J and 3K), suggesting that APOBEC3A expression in TLS may reflect the initiation of ESCC. We further analyzed the correlation between APOBEC3A expression and various immune cell populations within TLS, as well as its correlation with key immune regulatory genes. The results showed that APOBEC3A exhibited a significant positive correlation with plasma cells (R = 0.77, p = 0.021); among immune regulatory genes, APOBEC3A also displayed the strongest positive correlation with IGHM (a key marker of B cell activation and antibody secretion) (Figures S4B and S4C). These correlative findings provide initial clues to the potential regulatory mechanism of APOBEC3A, suggesting it may be involved in modulating B cell/plasma cell function within TLS. Taken together, these results suggest that the distinct regulatory functions of macrophages, neutrophils, and TLS provide fundamental insights for understanding the complex cross-regulation of different immune cell types during the pathological process of ESCC initiation.
Transcriptomic landscape in TME reveals characterization of transition from early ESCC to advanced stages
Understanding how esophageal precancerous lesions evolve into invasive or metastatic tumors may have profound significance for improving ESCC prognosis. In this study, we also intended to identify the underlying molecular regulatory mechanisms of EP during the process from early-stage (ESPL and non-mESCC) to advanced-stage (mESCC and mLNs) ESCC (Figure 4A; Table S3); several cancer-progression-associated pathways, such as PI3K/AKT signaling pathway and cell migration, adhesion, and proliferation processes, were significantly upregulated in advanced tumors (Figure 4B). Notably, the PI3K/AKT signaling pathway was continuously upregulated during the tumor initiation process, and additional genes, such as cyclin D1 (CCND1) and laminin subunit beta-1 (LAMB1), in this pathway were further upregulated during the advanced stage (Figure 4C), suggesting that the continuous activation of this pathway is crucial for the initiation and development of ESCC. Notably, in contrast to the genes activated in tumor initiation stage, certain genes related to nutrient response, such as OGT and TFRC, were only activated in the advanced stage. Moreover, the reduction of inflammatory and innate immune responses in the advanced stage suggests that they are primarily activated in the tumor initiation stage but inhibited during tumor progression. Notably, genes involved in neutrophil aggregation and activation, such as interleukin (IL-18) gene, were downregulated in the advanced stage. Based on mIF experimental verification, we determined that the protein expression of CCND1 and LAMB1 from the PI3K/AKT pathway significantly increased during tumor progression stages (all, p < 0.01), while IL-18 significantly decreased (p < 0.05) (Figures 4D and 4E).Figure 4. Characteristic changes in transcription patterns during the progression of ESCC(A) The volcano plot shows significant differences in genes between the early (ESPL and non-mESCC) and advanced (mESCC and mLN) stages of ESCC.(B) Pathway enrichment analysis of differentially expressed genes in the early and advanced stages of ESCC.(C) The gene expression levels associated with significantly enriched pathways throughout the occurrence and development of ESCC.(D) mIF staining of CCND1, LAMB1, and IL-18 on ESCC tissues. Scale bar, 200 μm.(E) Comparison of the percentages of CCND1, LAMB1, and IL-18-positive cells on ESCC tissues. Box plots show the median and interquartile range, whiskers extend to 1.5 × IQR. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, p values were calculated using a two-sided Wilcoxon rank-sum test.(F) The heatmap of gradient-changed DEGs based on pseudotime and ESCC progression of EP.
For a systematic elucidation of tumor development, we subsequently conducted pseudotime analysis to simulate the evolutionary trajectories of ESCC. The analysis based on EP cells showed a high correlation between their transitions and the development stages of ESCC (Figure 4F). We then took the intersection of the advanced-stage-specific gene set and pseudotime-associated gene set, yielding nine overlapping genes (Figure S5A). We further assessed the association between the expression profiles of these genes and clinical outcomes based on The Cancer Genome Atlas (TCGA)-ESCC dataset and found that only OGT exhibited a significant association with poor prognosis (Figure S5B). In addition, we compared OGT with other established stage-specific markers (e.g., SPP1 and MMP9) to clarify its prognostic utility. The results showed that high OGT expression was significantly associated with poorer patient prognosis (hazard ratio [HR] = 1.66, 95% confidence interval [CI] = 1.04–2.7, p = 0.034) (Figure S5C), indicating that high OGT expression may serve as an independent risk factor for ESCC progression. Based on the public ESCC scRNA-seq dataset, we confirmed that OGT was specifically activated in tumor cell populations during the advanced stages of ESCC (Figure S5D). To explore the broader prognostic value of OGT beyond ESCC, we stratified patients from other TCGA tumor cohorts based on OGT expression levels and analyzed their DFS. The findings revealed that in several TCGA cohorts—including uterine corpus endometrial carcinoma (UCEC), uveal melanoma (UVM), kidney renal papillary cell carcinoma (KIRP), and mesothelioma (MESO)—patients with high OGT expression exhibited significantly reduced DFS (Figure S5E). This confirms that OGT can serve as a prognostic marker for DFS across multiple cancer types, highlighting its potential as a universal prognostic factor. Collectively, these findings establish OGT as a factor with profound functional and clinical significance in advanced ESCC, underscoring the need for further in-depth investigation into its biological mechanisms that drive ESCC progression.
OGT can enhance the migratory and invasive capabilities of ESCC cells in vivo and in vitro
To further elucidate the biological function of OGT in ESCC, we first conducted immunohistochemical analysis (IHC) for detecting the expression level of OGT based on 76 pairs of ESCC and adjacent non-cancerous tissues (Figure 5A). The results showed significantly increased OGT expression in ESCC tissues compared to that in adjacent non-cancerous tissues (Figure 5B). Subsequently, the clinical and pathological data of samples were analyzed. A statistically significant correlation was observed between OGT expression level and the pathologic T stages as well as TNM stages in patients with ESCC, while no significant correlations were found between OGT expression and the patients’ age, gender, smoking history, alcohol consumption history, or tumor size (Table S5), indicating that OGT might be an important risk factor involved in ESCC progression.Figure 5OGT promotes ESCC progression through regulating proliferation, migration, invasion, and apoptosis in vitro and in vivo(A) Representative images of immunohistochemical staining for OGT in cancer and adjacent tissues, magnified at 200× and 400×; scale bars, 100 μm.(B) Paired comparison of OGT IHC scores in tumor and peritumor tissues. p value was calculated using a two-sided Wilcoxon signed-rank test (paired samples, n = 76).(C) Representative WB images of OGT and O-GlcNAc level in all transient transfection groups, including knockdown in KYSE30 and KYSE450 (transfected with siNC, siOGT#1, or siOGT#2) and overexpress in KYSE150 and KYSE410 (transfected with oeVector, oeOGT, or oeOGT and treated with OSMI-1). Images are representative of three independent experiments (n = 3).(D) OGT and O-GlcNAc level in lentiviral-mediated knockdown KYSE450 and overexpress KYSE150. Representative protein images of three independent experiments were shown (n = 3).(E and F) Representative images and quantification of plate clone formation assay of lentiviral-mediated knockdown KYSE450 and overexpress KYSE150 cell lines. Data are mean ± SD from three independent experiments (n = 3). p values were calculated using one-way ANOVA.(G) Cell proliferation measured by the CCK8 assay and relative cell proliferation quantified by OD value at 450 nm of KYSE30, KYSE450, KYSE150, and KYSE410 cell lines. Data are presented as mean ± SD from three independent experiments (n = 3). p values were calculated using two-way ANOVA with Tukey's multiple-comparisons test.(H) Representative images of cell migration and invasion assays in the KYSE30, KYSE450, KYSE150, and KYSE410 cell lines (n = 3). Scale bars, 200 μm.(I) Stably transfected shNC/shOGT KYSE450 cell subcutaneously injected into nude mice. Representative image of xenograft tumors and tumor volume from day 5 to day 30. Data are presented as mean ± SD, n = 5 mice per group. p values were calculated using two-way repeated-measures ANOVA with Sidak's multiple-comparisons test.(J) Bioluminescence imaging of lung metastatic foci at the 7th week in a lung metastasis model. Luciferase activity is measured in photons per cm2 per second per steradian (p/s/cm2/sr).(K) Representative images and quantitative analysis of metastasis nodules on the lung surface. Arrowheads denote the metastasis nodules on the lung surface.(L) Representative images of HE staining of lung metastasis and quantitative analysis of lung metastasis area. Arrowheads denote the metastasis nodules. Scale bars, 200 μm. (J–L) Data are presented as mean ± SD, n = 5 mice per group. p values were calculated using a two-sided Wilcoxon rank-sum test.(M) The results of enriched pathways affected by OGT expression detected across different omics. In the RNA-seq dataset, siOGT cells were defined as the OGT low group, while siNC cells were defined as the OGT-high group. In both the proteomic and glycoproteomic datasets, oeVector cells were defined as the OGT-low group, and oeOGT cells were defined as the OGT-high group. Pathways activated in the OGT-high group are marked in green, while pathways inhibited are marked in red.(N) Comparison of apoptosis rate between KYSE450 siNC and siOGT groups. Data are presented as mean ± SD, n = 3 per group. p values were calculated using unpaired Student's t test.(O) Flow cytometry analysis of cell cycle distribution of KYSE450 cells across siNC and siOGT. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001. NC, negative control group.
We then tested the mRNA level of OGT gene in normal esophageal HEEC cell line and four esophageal squamous carcinoma cell lines (KYSE30, KYSE150, KYSE410, and KYSE450) by RT-qPCR (Figure S6A). Next, transient and stable overexpression or knockdown models were established, while the overexpression group was further subjected to the small-molecule OGT inhibitor OSMI-1(30 μM), to preliminarily evaluate the functional role of OGT in ESCC cell behavior. The transfection efficiency was confirmed by western blot (WB) analysis, also showing the level of O-glycosylation is positively associated with OGT, and was markedly downregulated upon OSMI-1 treatment (Figures 5C and 5D). The malignant phenotypes of ESCC cells, assessed by plate clone formation, Cell Counting Kit-8 (CCK8), Transwell, and wound-healing assays, were significantly suppressed by OGT knockdown. In contrast, OGT overexpression markedly enhanced these malignant phenotypes; however, this effect was substantially attenuated upon treatment with OSMI-1 (Figures 5E–5H, S6B, and S6C). These results also demonstrated that OGT-mediated enhancement of cell motility is dependent on its enzymatic activity. To investigate the role of OGT in promoting tumor growth and metastasis in vivo, two types of xenograft mouse models were established using KYSE150 and KYSE450 cells. In the KYSE450 subcutaneous model, OGT knockdown markedly suppressed tumor growth, with significantly smaller tumor volumes compared to controls (Figure 5I). In the lung metastasis model, at the 7th week after injection, mice of oeOGT group exhibited significantly stronger lung bioluminescent signals (Figure 5J), more surface nodules (Figure 5K), and greater metastatic area on H&E sections (Figure 5L), compared to controls, indicating an enhanced metastatic potential. Taken together, evidence from both in vitro and in vivo experiments indicates that OGT plays a promotive role in the progression and metastasis of ESCC, underscoring its potential as a pivotal molecular target in ESCC research.
To further investigate the molecular mechanism of OGT in ESCC, we first performed whole-transcriptome sequencing on OGT-knockdown and corresponding control cell lines, identifying 568 differentially expressed genes (DEGs), of which 375 were upregulated and 193 were downregulated in the OGT-knockdown group (Figure S6D; Table S3). Concurrently, we constructed a co-expression network based on whole-transcriptome DSP data to screen genes highly correlated with OGT, leading to 458 significantly OGT-associated DEGs (Figure S6E). Enrichment analysis revealed that pathways related to keratinocyte proliferation, cell adhesion, and apoptotic processes were activated in the OGT-low expression group, whereas pathways including cell cycle, DNA replication, and tumor-progression-associated pathways such as Janus kinase-signal transducers and activators of transcription (JAK-STAT) and Wnt signaling pathways were significantly enriched in the OGT-high expression group, with these findings further validated by proteomic profiling analyses (Figure 5M). Considering OGT’s extensive involvement in protein modification, we performed quantitative O-GlcNAc glycoproteomic analysis using OGT-overexpressing and Vector ESCC cells, a mass-spectrometry-based approach enabling identification of O-GlcNAcylated substrates modulated by OGT expression. By integrating proteomic and glycosylation mass spectrometry results, we confirmed the activation of cell-proliferation-related pathways in the OGT-high group and enrichment of cell apoptosis in the OGT-low group. Based on flow-cytometry-based functional assays, we observed that OGT knockdown increased apoptosis rates in KYSE450 cells and altered cell-cycle distribution, characterized by elevated G2/M phase proportions and decreased S phase proportions (Figures 5N, 5O, S6F, and S6G). Given that several OGT-associated pathways can also be driven by constitutive PI3K/AKT activation, we further investigated the functional crosstalk relationship between OGT and the PI3K/AKT pathway. We found that increased O-GlcNAcylation induced by OGT overexpression was associated with elevated AKT Ser473 phosphorylation, which was effectively suppressed by the PI3K inhibitor LY294002 or the OGT inhibitor OSMI-1. Notably, neither total nor phosphorylated PI3K, nor total AKT protein levels, were altered across conditions (Figure S6H). These results indicate that OGT influences AKT phosphorylation through O-GlcNAcylation downstream of PI3K, potentially impacting multiple AKT-dependent pathways, including JAK-STAT and Wnt signaling. In summary, these findings demonstrate that OGT expression plays a critical role in ESCC progression by promoting multiple pro-tumor pathways and inhibiting apoptotic pathways, highlighting OGT as a potential therapeutic target in esophageal squamous cell carcinoma.
Characterization of TME remodeling and TLS that promotes tumor invasion
We next examined how tumor ecosystem remodeling occurs during ESCC progression. Comparison of the proportion of different cell subtypes revealed highly dynamic changes in cell abundance during the early and advanced stages of ESCC (Figure 6A). A substantial decrease in mast cells, neutrophils, and endothelial cells and an increasing trend in macrophages were observed in the advanced stage (Figures 6B and S7A). To further understand the changes and interactions within the TME during ESCC progression, we conducted differential analyses on ST, MC, and NC compartments (Figures 6C–6E; Table S3). Pathway enrichment analysis based on their respective DEGs revealed that in ST compartment, PI3K/AKT, angiogenesis, and cell migration pathways were more active in the advanced stage (Figure S7B). In MC compartment, positive regulation involving neutrophils and T cells was shifted toward enhanced cholesterol metabolism and extracellular space activities in the advanced stage (Figure S7B). In the NC compartment, although significant involvement of immune responses (such as B cell activation and leukocyte migration) was observed in the early stages, suppressed immune activity and a shift toward epidermal cell differentiation and glucose transmembrane transport regulation was observed in the later stages of ESCC (Figure S7C). Further, to understand how the DEGs involved in these pathway changes interact within different compartments, we screened the aforementioned DEGs in the CellChat database, identifying cell-cell interaction (CCI) pathways directly involved in the advanced stages of ESCC (Figure 6F; Table S6). These cell-cell interactions between different compartments occurred via ACTIVIN [INHBA->(ACVR1B + ACVR2)], CHEMERIN (RARRES2->CMKLR1), and PERIOSTIN [POSTN->(ITGAV+ITGB5)] pathways (Figure 6G). Further analysis of CCI pathways across normal esophageal tissue, early-stage ESCC, and advanced-stage ESCC samples uncovered stage-specific activation patterns of key signaling axes (Figure 6H). ACTIVIN-, CHEMERIN-, and PERIOSTIN-mediated interactions showed no significant signal intensity in the normal stage, whereas all three pathways were activated in early-stage ESCC, manifesting as intercellular crosstalk between ST and EP/TLS; among these, the POSTN and ACTIVIN pathways exhibited further enhanced signal strength in the advanced stage. Notably, ACTIVIN-, CHEMERIN-, and PERIOSTIN-dependent CCIs between ST and MC/NC were exclusively detected with significantly increased signal intensity in advanced-stage ESCC and remained inactive in the normal and early stages. Additionally, evaluation of immune-regulatory CCI pathways revealed that CXCL13 expression was consistently upregulated in TLS throughout ESCC progression, and CXCL13 engaged in robust intercellular communication with MC, NC, and ST via the receptors CXCR5, CXCR3, and ACKR1 (Figure S7C). These findings collectively demonstrate the progressive rewiring of CCI networks during ESCC progression, with stage-specific induction and enhancement of stromal-epithelial/TLS and stromal-immune interactions.Figure 6. Changes in TME during the progression of ESCC(A) Abundance of 17 cells estimated by SpatialDecon algorithm in non-EP compartments between the early (ESPL and non-mESCC) and advanced stages (mESCC and mLN) of ESCC.(B) Comparison of proportion of presented cells between the early and advanced stages of ESCC. Box plot shows the median and interquartile range, whiskers extend to 1.5 × IQR. p values were calculated using two-sided Wilcoxon rank-sum test. ∗p < 0.05, ∗∗∗p < 0.001, ∗∗∗∗ p < 0.0001.(C) Volcano plot showing the DEGs between early and advanced stages in ST compartment.(D) Volcano plot showing the DEGs between early and advanced stages in MC compartment.(E) Volcano plot showing the DEGs between early and advanced stages in NC compartment.(F) Cell-cell interactions based on significant ligand-receptor pairs in advanced stages of ESCC.(G) Interaction relationships of the ACTIVIN, CHEMERIN, and PERIOSTIN pathways among compartments in advanced ESCC.(H) Cell-cell interactions between ST compartments and other compartments via the ACTIVIN, CHEMERIN, and PERIOSTIN pathways during ESCC initiation and progression.(I) The four-quadrant diagram showed the significant differences in gene expression patterns among three different types in the TLS during the ESCC process. Genes showing continuous increase were marked in red, while genes showing continuous decrease were marked in blue.(J) Volcano plot showed significant differentially expressed genes between the TLS of tumor and the lymph follicles of lymph nodules.
A recent study reported that TLS change plays an important role in the development of ESCC.26 To explore the changes in TLS, which are a crucial part of the TME during ESCC progression, we compared DEGs in TLS throughout tumor progression. We identified three distinct gene expression patterns (Figure 6I). For example, chemokine (C-X-C motif) ligand 13 (CXCL13), marginal zone B- and B1-cell-specific protein (MZB1), and H4 clustered histone 6 (H4C6) were consistently and significantly overexpressed during tumor initiation and progression. CXCL13 is a major chemokine involved in recruiting B cells and T follicular helper (Tfh) cells, validating the formation of TLS during tumor progression. Conversely, chemokine (C-C motif) ligand 19 (CCL19) and alpha-2-macroglobulin (A2M) were significantly suppressed during this process. Since these immune-related genes play a critical role in inhibiting tumor cell migration and metastasis, their suppressed expression indicates a compromised ability of immune cells to eliminate and inhibit tumor metastasis, leading to a higher likelihood of migration in advanced stages. Correlation analysis between genes associated with TLS dysfunction and key exhausted T cell (Tex) markers (e.g., PDCD1, LAG3, and TOX) revealed that genes with upregulated expression in dysfunctional TLS exhibited a positive correlation with the Tex markers PDCD1 and TOX (Figure S7D). To further dissect the potential drivers of TLS dysfunction, we computed the dysfunction signature score for each TLS and its correlation with the signature scores of dysregulated signaling pathways in late-stage EP and ST compartments. Analysis of samples harboring both EP and TLSs, or ST and TLS compartments, revealed that dysregulated pathways from EP compartments exhibited a relatively strong positive correlation with TLS dysfunction, whereas those from ST compartments showed weak or negative correlations (Figures S8A and S8B). When focusing on TLS compartments, we observed a significant positive correlation between the EP-associated “response to nutrient” signature score and TLS dysfunction (R = 0.68, p = 0.014) (Figure S8A). Notably, compared with other genes in the “response to nutrient” pathway, TFRC and OGT genes showed the strongest correlations with TLS dysfunction (Figure S8C). These findings suggest that during late-stage ESCC, TLS dysfunction may be driven by aberrant “response to nutrient” signaling pathways in tumor epithelial compartments.
Regarding the process of ESCC metastasis, our analysis of DEGs in TLSs and lymph follicles within lymph nodes revealed notable findings (Figure 6J; Table S3). The significant increase of CCL21, a chemokine that recruits T cells and dendritic cells, and TRAF-family-member-associated nuclear factor κB (NF-κB) activator (TANK), a gene involved in the NF-κB pathway, suggest an enhanced immune response in lymph follicles. However, the overexpression of exocyst complex component 3-like protein 4 (EXOC3L4) and tenascin-C (TNC), genes associated with tumor metastasis, indicates that tumor cells in mLNs strongly resist the immune system. Conversely, in TLSs, while some genes regulating immune functions remain significantly elevated, the expression of cytokines related to immune cell recruitment was suppressed. This implies that in the advanced stages of tumor progression, TLSs are unable to recruit new immune cells, thereby losing their ability to effectively inhibit tumor cells.
Discussion
Exploring the onset and progression of ESCC can help identify key factors driving tumorigenesis and metastasis and provide deeper insights into the TME changes that occur during tumor evolution. To thoroughly understand the entire processes, it is essential to analyze the alterations both in the epithelial cells and their surrounding environment. In this study, we conducted a systematic digital whole-transcriptome spatial molecular analysis of normal, ESPL, ESCC, and mLN tissues. By focusing on both epithelial cells and TME, we comprehensively characterized the spatial details and molecular differences during ESCC initiation and progression. The application of DSP technology provides a distinct avenue for cancer evolution research, offering insights that other technologies cannot provide. We validated our findings using mIF, which confirmed the changes associated with molecular features and immune cells. Functional studies on OGT revealed its protumorigenic role in ESCC. Inhibiting OGT overexpression significantly reduced ESCC cell proliferation and invasion, highlighting its potential as a therapeutic target.
During the progression from normal tissue to ESPL, significant activity occurs in the epidermis development pathway. Genes such as SOX4 and KRTDAP, which exhibit embryonic-like characteristics,27^,^28 were overexpressed, enhancing the stemness of epithelial cells. As ESPL progresses to early ESCC, keratin differentiation is continuously suppressed, reflecting the transformation of esophageal cells into squamous cells. The development of ESCC is not only driven by changes in the epithelial compartment but is also closely linked to the TME.10^,^29^,^30 Detailed investigation of the TME revealed an increase in B cells during the ESPL stage, coinciding with the formation of TLS. Additionally, significantly increased APOBEC3A expression within TLS during tumor onset suggests its role in tumor escape and poor prognosis. Meanwhile, a notable decrease in the number of CD4^+^ and CD8^+^ T cells along with a continuous increase in regulatory T cells (Tregs) in non-mESCC stages indicate the establishment of an immunosuppressive environment during tumor initiation. Furthermore, the gene expression patterns in MC and NC compartments, which suppress immune responses and programmed cell death, highlight their involvement in tumor formation.
To identify key factors driving the development of ESCC, we compared the early and advanced stages of ESCC progression. Our analysis revealed that pathways related to PI3K/AKT signaling, nutrient response, and cell migration were consistently activated. We validated these findings using mIF, confirming significantly upregulated CCND1 and LAMB1 within these pathways. A recent study published in Cancer Cell employed the 10× Xenium platform to investigate ESCC evolution at single-cell spatial resolution, highlighting epithelial-CAF interactions and the early activation of NOTCH signaling.31 Prompted by this work, we reanalyzed our DSP data with a shifted focus to epithelial and stromal regions, where CAFs are primarily localized. This supplementary analysis revealed that the JAG1-NOTCH axis is indeed activated in EP-ST interactions during the early stage of ESCC, with its signal intensity further increasing in the advanced stage (Figure S7E). This finding provides complementary evidence for the role of NOTCH signaling in epithelial-stromal crosstalk during ESCC progression, alongside our previously reported activation of the PI3K/AKT pathway. Notably, a key advantage of the DSP platform for our study lies in its ability to perform unbiased whole-transcriptome profiling following the definition of user-specific AOIs. This stands in contrast to the 10× Xenium platform, which depends on pre-designed gene panels and thus restricts detection to a fixed set of transcripts. For our research focus, the transcriptome-wide coverage of DSP enabled us to capture a more comprehensive spectrum of biological functions.
By integrating gene sets associated with advanced-stage ESCC and evolutionary trajectories, together with the clinical significance of candidate genes, we focused our investigation on the regulatory mechanisms of the OGT gene. In our in vivo and in vitro experiments, we further confirmed its tumor-promoting function, suggesting that OGT, through its glycosylation function, may serve as a key driver in ESCC development, especially in metastasis and invasion. As the central enzyme regulating O-GlcNAcylation, OGT has been implicated in multiple cancers, and elevated OGT expression enhances O-GlcNAcylation, thereby influencing tumor glucose metabolism, cell proliferation, invasion, angiogenesis, and resistance to therapy. Consistent with previous studies,32^,^33^,^34 following glycoproteomic profiling further suggests that OGT promotes ESCC progression by increasing the O-GlcNAcylation of numerous proteins enriched in cell-cycle, DNA-replication-related pathways. The regulatory effect of OGT on phosphorylated AKT has also been confirmed. In vitro flow cytometry assay demonstrated concordant findings. The aforementioned results and the inhibitory activity of OSMI-1 highlight OGT as a promising therapeutic target in ESCC.
However, given that OGT is a ubiquitous and essential enzyme, concerns remain regarding its feasibility as a therapeutic target. The essential nature of OGT is well established by genetic studies showing that complete knockout is embryonically lethal, underscoring both its critical biological roles and the inherent risk that global inhibition could perturb normal physiology.35^,^36 This has raised concerns that direct inhibition might lead to adverse systemic effects, owing to the enzyme’s involvement in modifying thousands of substrates and regulating key cellular processes.37^,^38 Several small-molecule OGT inhibitors, such as OSMI-1, have demonstrated antitumor effects in cell lines and animal models; however, none have yet advanced to clinical application. Therapeutic targeting of OGT faces several challenges, including off-target toxicities,37^,^39^,^40 poor cell permeability, and low aqueous solubility,36^,^39^,^41 which limit the effectiveness of OGT inhibitor in vivo. However, various research papers have approached these obstacles with different strategies: advanced chemical tools such as glycoproteomics techniques and chemoenzymatic labeling were employed to better define OGT substrate specificity.37^,^42 Nanotechnology has been explored for improving drug delivery, overcoming poor solubility and permeability issues.41 Combining OGT inhibitors with other antitumor agents offers a strategy to reduce monotherapy dosage and mitigate risks. Mechanistically, OGT-mediated O-GlcNAcylation intersects with key signaling pathways (NF-κB, JAK/STAT, and MAPK/ERK) and tumor immune regulation, supporting rational combination therapy. Notably, OGT inhibitors may synergize with PD-1/PD-L1 blockade to improve the tumor immune microenvironment43^,^44 or with MAPK/ERK and PI3K inhibitors to suppress proliferative and metabolic signaling.45^,^46 In addition, combining with chemotherapy could enhance chemosensitivity by intensifying metabolic stress.43 Though many challenges remain, these efforts highlight a positive research trajectory, and the potential of OGT as a therapeutic target continues to grow. In the future, we will continue to carry out more comprehensive studies to clarify the underlying mechanisms and advance its clinical translation.
During TME remodeling with ESCC progression, we observed that Tregs maintained a relatively stable population, whereas NC exhibited an opposite trend. In terms of pathway enrichment, the positive regulation of neutrophil and T cell signaling in MC decreased, while NC showed a more pronounced role in immune suppression. Furthermore, ST cells interacted with components of the TME, playing a crucial role in angiogenesis and tumor metastasis.47 Our analysis of cell interactions revealed that ST cells had enhanced communication with MC and NC via the ACTIVIN, CHEMERIN, and PERIOSTIN pathways. This interaction further shapes the TME, contributing to the progression and metastasis of ESCC.48^,^49^,^50^,^51^,^52^,^53 For TLS, we not only compared their molecular characteristics during ESPL and advanced ESCC but also contrasted them with lymphoid follicles in mLNs. Differences in cytokine profiles resulted from changes in TLS status.54^,^55^,^56 Our study showed decreased CXCL21 expression in TLS during the advanced stages of ESCC, suggesting a loss of immune cell recruitment capability that leads to impaired suppression of tumor cells. This indicates that TLS assessment should not be limited to simply determining their presence.26 Instead, evaluating their functional state through expression patterns is essential for a comprehensive understanding of their role in cancer metastasis. TLSs are organized aggregates of adaptive immune cells that arise in non-physiological, non-lymphoid tissues. Analogous to secondary lymphoid organs (SLOs) in their organization, classical TLSs consist of two principal compartments: a B-cell zone forming a follicle-like structure at the center and a surrounding T-cell-rich zone.57^,^58 These structures typically localize at the interface between tumor parenchyma and adjacent stroma, or at the invasive margin of tumors, and are considered pivotal sites for anti-tumor immune responses. TLS presence often correlates with intensified local immune activity and improved clinical outcomes, playing roles in promoting anti-tumor immunity, activation of T and B cells, antibody production, and modulation of the immune microenvironment.59^,^60
Consistent with previous findings, our current study observed that during early tumorigenesis, expression of APOBEC3A within TLS increases as the tumor progresses and is positively correlated with the B cell activation marker IGHM. In contrast, in advanced ESCC, TLSs display diminished expression of CXCL21, indicating impaired immune cell recruitment and consequent weakened anti-tumor capability, whereas lymph nodes continue to recruit immune cells. This observation suggests a stage-specific immunological strategy: in early-stage tumors, preserving TLS functionality is critical, whereas in later stages, therapeutic focus might need to shift toward enhancing lymph-node-mediated immunity to compensate for TLS dysfunction. Moreover, APOBEC3A may serve as a molecular marker initiating early immune surveillance, and the TLS-APOBEC3A activity axis could be leveraged as a strategic target for early diagnosis, risk stratification, and immunoprevention. It is also worth noting that while our findings suggest TLS dysfunction during late-stage ESCC may be driven by aberrant “response to nutrient” signaling in EP compartments, these observations are based solely on correlative analyses and thus remain speculative. Definitively delineating the detailed mechanisms underlying TLS dysfunction will require more rigorous, mechanistic experimental investigations. These represent an important direction that merits further exploration in our next stage of research.
In summary, our study reveals the collaborative roles of epithelial cells and the TME during the initiation and progression of ESCC. These coordinated changes contribute to the development and metastasis of ESCC. Our observations provide a systematic understanding of the interactions between tumors and the immune microenvironment.
Limitations of the study
Despite these important discoveries, we acknowledge the potential limitations in this study. First, all ESPL cases analyzed are classified as high-grade intraepithelial neoplasia, and no patients with low-grade intraepithelial neoplasia were included to study the occurrence of esophageal cancer. Early molecular and cellular events from low- to high-grade lesions could not be directly assessed, limiting the immediate translation of the identified molecular or spatial features to ultra-early risk stratification or biomarker development. Second, the validation cohort for mIF corresponds to the same patient cohort utilized for DSP. Although these two approaches provide complementary and orthogonal biological information, the absence of an independent external cohort could potentially limit the broader generalizability of our findings. We intend to broaden the validation cohort to further corroborate the data in subsequent investigations.
Resource availability
Lead contact
Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Zhao Liu ([email protected]).
Materials availability
This study did not generate new unique reagents.
Data and code availability
- •The sequencing data have been deposited in the Genome Sequence Archive for Human repository (https://cncb.ac.cn): HRA008846.
- •This article does not report original code.
- •Any additional information required to reanalyze the data reported in this work paper is available from the lead contact upon request.
Acknowledgments
We thank all the study participants. The graphical abstract was created in part using BioRender.com. This study was funded by National Natural Science Foundation of China (grant nos. 82573478 and 82303066), Hebei Natural Science Foundation (grant nos. H2025206822 and H2023206006), Postdoctoral Fellowship Program of CPSF (grant no. GZB20230189), the Project of Funds for the Centre-Guide-Local Development of Science and Technology by Provincial Department of Science and Technology (grant no. 236Z7746G), Support Plan for Research and Innovation Team (Class A) of the Fourth Hospital of Hebei Medical University (grant no. 2023A02), and Hebei Key Laboratory of Accurate Diagnosis and Comprehensive Treatment of Esophageal Cancer (grant no. SZX202304).
Author contributions
Conceptualization, Z.T. and L.Z. Methodology, M.Y., H.L., and J.W. Investigation, Z. Liu. and Z. Li. Visualization, W.Z., Z.L., and M.Z. Supervision, Z.T. and L.L. Writing—original draft, H.L. and Z.L. Writing—review & editing, N.L., L.Z., H.L., and Z.T.
Declaration of interests
The authors declare no competing interests.
STAR★Methods
Key resources table
REAGENT or RESOURCESOURCEIDENTIFIERAntibodiesRecombinant Rabbit monoclonal anti-OGTHUABIOCat# ET7107-17; RRID: AB_3070732Mouse monoclonal anti- O-Linked N-AcetylglucosamineAbcamCat# ab2739; RRID: AB_303264Rabbit Polyclonal anti-beta ActinServicebioCat# GB11001; RRID: AB_2801259Anti human PancytokeratinAbcamCat# ab234297; RRID:AB_ 2895302Anti human CD4AbcamCat# ab133616; RRID: AB_2750883Anti human CD8Thermo Fisher ScientificCat# MA1-80231; RRID: AB_929437Anti human CD20Cell Signaling TechnologyCat# 48750S; RRID:AB_3107071Anti human FOXP3Cell Signaling TechnologyCat# 98377; RRID: 2747370Anti human CD68AbcamCat# ab280860; RRID: AB_2801637Anti human CD163AbcamCat# ab182422; RRID: 2753196Anti human CD86AbcamCat# ab220188; RRID: 3676334Cyclin D1 Recombinant Rabbit Monoclonal AntibodyHUABIOCat# ET1601-31; RRID: AB_3069614Laminin beta 1 Recombinant Rabbit Monoclonal AntibodyHUABIOCat# ET1703-14; RRID: AB_3070371IL-18 Recombinant Rabbit Monoclonal AntibodyHUABIOCat# HA721536; RRID: AB_3072652AKT1/2/3 Recombinant Rabbit Monoclonal AntibodyHUABIOCat# ET1609-51; RRID: AB_2940862Phospho-AKT (S473) Recombinant Rabbit Monoclonal AntibodyHUABIOCat# ET1607-73; RRID: AB_2940863PI 3 Kinase p85 alpha Recombinant Mouse Monoclonal AntibodyHUABIOCat# HA601206; RRID: AB_3071925Phospho-PI3 Kinase p85 (Tyr458)/p55 (Tyr199) AntibodyCell Signaling TechnologyCat# 4228; RRID: 659940Bacterial and virus strainsLentivirus: shRNA targeting OGTGenePharmaCat# D01001Lentivirus: control shRNAGenePharmaCat# D03001Lentivirus: OGT overexpressionGeneralbiolCat# V1342936Lentivirus: blankGeneralbiolCat# V1342935Biological samplesESCC patient samplesThis paperN/AChemicals, peptides, and recombinant proteinsOSMI-1TargetmolCat# T16409LY294002CaymanCat# 70920D-Luciferin (potassium salt)Cayman chemicalCat# 14681PuromycinCayman chemicalCat# 13884Critical commercial assaysGP-transfect-MateGenePharmaCat# G04008Opal 7-Color KitAkoya BiosciencesN/ASuperbrilliant® 6 min 1st Strand cDNA Synthesis KitZhongshi Gene TechnologyCat# ZS-M14003Superbrilliant® 2×ZAPA3G SYBR Green qPCR MixZhongshi Gene TechnologyCat# ZS-M13002Cell Counting Kit-8MedChemExpressCat# HY-K0301PE Annexin V Apoptosis Detection Kit IBD PharmingenCat# 559763PI/RNase Staining BufferBD PharmingenCat# 554656Deposited dataRaw sequencing dataThis paperGenome Sequence Archive for Human (https://cncb.ac.cn): HRA008846TCGA datasetshttps://portal.gdc.cancer.govN/AAberrant epithelial cell interaction promotes esophageal squamous-cell carcinoma development and progressionLin et al.14https://doi.org/10.1038/s41392-023-01710-2Experimental models: Cell linesKYSE150 human ESCC cell lineHaixing BiosciencesTCH-C236KYSE410 human ESCC cell lineHaixing BiosciencesTCH-C458KYSE450 human ESCC cell lineServicebioSTCC11902PKYSE30 human ESCC cell lineServicebioSTCC11903PExperimental models: Organisms/strainsMouse: BALB/c-nuHFK BioscienceRRID:13001AMouse: NOD SCIDHFK BioscienceRRID:13002AOligonucleotidesPrimers for RT-qPCRThis paperN/ARecombinant DNAPlasmids and RNA interference sequencesThis paperN/ASoftware and algorithmsGraphPad Prism 9Graphpad softwarehttps://www.graphpad.comImageJhttps://imagej.net/ij/R (v4.0.0)https://www.rproject.orgGEPIA2Zhang Lab61https://gepia2.cancer-pku.cn/Seurat (v5.3.0)Hao et al.62https://satijalab.org/seurat/edgeR (v4.6.2)Smyth et al.63https://bioconductor.org/packages/release/bioc/html/edgeR.htmlclusterProfiler (V4.8.3)Yu et al.64N/AGeoMx NGS pipelinenanostringhttps://nanostring.comGeoMx DSP systemnanostringhttps://nanostring.comComplexHeatmap (v2.24.0)Gu et al.65https://bioconductor.org/packages/release/bioc/html/ComplexHeatmap.htmlggplot2 (v3.5.2)Pedersen et al.66https://tidyverse.r-universe.dev/ggplot2Cibersort (v0.1.0)Alizadeh et al.67https://doi.org/10.1007/978-1-4939-7493-1_12SpatialDecon (v1.18.0)Dandher et al.68https://doi.org/10.18129/B9.bioc.SpatialDeconConsensusTME (v0.0.1)Miller et al.69https://doi.org/10.1158/0008-5472.CAN-18-3560SCORPIUS (v1.0.9)Cannoodt et al.70https://github.com/rcannood/SCORPIUSCellChat (v1.5.0)Jin et al.71https://github.com/sqjin/CellChatSTRINGSTRING Consortium72https://cn.string-db.orgcorrplot (v0.95)Wei et al.73https://github.com/taiyun/corrplotsurvival (v3.8-3)Therneau et al.74https://cran.r-project.org/web/packages/survival/index.htmlsurvminer (v0.5.0)Kassambara et al.75https://cran.r-project.org/web/packages/survminer/index.htmlForm (v2.4.4)Akoya Bioscienceshttps://www.pubcompare.ai/product/zzPiCZIBPBHhf-iFwRy1/
Experimental model and study participant details
Mouse models
6 weeks male BALB/c-nu and NOD SCID mice were purchased from the Beijing Huafukang Bioscience Co., Ltd. All mice were maintained in SPF-grade sterile conditions, and all procedures were approved by the Institutional Animal Care and Use Committee of the 4th hospital of Hebei Medical University. (Approval No. IACUC-4th Hos Hebmu-20250529).
Human samples
Human samples were collected at the Fourth Hospital of Hebei Medical University. ESCC Patients who underwent surgical treatment at the Fourth Hospital of Hebei Medical University between 2022 and 2023 were screened and selected, including 6 cases in the precancerous lesion group, 6 cases in the early cancer group (T1aN0M0 or T1bN0M0 stage), and 6 cases in the advanced cancer group (T3N1M0, T3N2M0, or T3N3M0 stage). Other 76 pairs of ESCC and normal tissues were collected to evaluate the expression levels of OGT. Demographic and clinical characteristics of the patients, including age and gender, are summarized in Tables 1 and S5. All patients were histopathologically diagnosed with ESCC or ESPL. None of the patients had a history of autoimmune disease, chronic inflammatory disorders, or immunodeficiency. At the time of sample collection, patients were considered immunocompetent. None of the patients had received chemotherapy, radiotherapy, targeted therapy, or immunotherapy prior to tissue collection. The requirement for written informed consent was waived by the ethics committee because the study involved the use of archived diagnostic tissue specimens, was retrospective in nature, and posed minimal risk to participants. This study was approved by the Medical Ethics Committee of the Fourth Hospital of Hebei Medical University (approval number 2024KS224) and carried out under the World Medical Association Declaration of Helsinki.
Cell culture
Four human esophageal cancer cell lines, including KYSE150, KYSE410, purchased from Haixing Biosciences (Suzhou, Jiangsu, China), and KYSE30, KYSE450, purchased from Servicebio (Hubei, Wuhan, China), were cultured. No mycoplasma contamination was observed during cell culture. All cell lines were cultured with RPMI-1640 medium (GIBCO, USA) containing 10% fetal bovine serum and penicillin/streptomycin antibiotics in a humidified incubator with 5% CO_2_ at 37°C.
Method details
Sample collection
All formalin-fixed paraffin-embedded (FFPE) samples were consecutively sliced to a thickness of 4–6 μm. The paraffin sections of the selected pathological tissues were subjected to routine H&E staining to ensure that the selected samples included the target area for spatial omics detection. A tissue microarray (TMA) with a diameter of 4 mm was made, consisting of 32 cores, including 6 cores for NE, 6 for ESPL, 6 for non-mESCC, 7 for mESCC, and 7 for mLN. The other 76 pairs of ESCC and normal tissues were collected to evaluate the expression levels of OGT by immunohistochemistry.
H&E staining
FFPE samples from 18 patients were collected. H&E staining was performed on all tissue sections to design TMA with target areas for spatial omics analysis. Standard procedures were followed for detection, including dewaxing the paraffin sections in water, staining with hematoxylin for cell nuclei, staining with eosin for cytoplasm, and dehydration and sealing for image collection. Lymphoid structures were selected based on H&E staining images.
Immunohistochemistry
FFPE sections were dewaxed in xylene (three changes, 5 min each) and rehydrated through graded ethanol (100%, 95%, 80%, 70%; 1 min each). Antigen retrieval was performed by heating slides in citrate buffer (pH 6.0) using a pressure cooker until boiling; after steam release, heating was continued for 2 min. The cooker was cooled under running water, and sections were rinsed three times in PBS. To block nonspecific binding, sections were incubated with 10% goat serum for 30 min at room temperature. Primary antibody against OGT (1:1000) was applied for 2 h in a humidified chamber, followed by three PBS washes. Sections were then incubated with an HRP-conjugated goat anti-mouse/rabbit IgG polymer for 30 min, washed, and visualized using 3,3′-diaminobenzidine (DAB). The reaction was stopped with water once optimal signal intensity was reached. Slides were counterstained with hematoxylin, dehydrated in graded ethanol, cleared in xylene, and mounted with neutral resin. IHC staining was independently evaluated by two pathologists, and discrepancies were resolved by joint review. Staining intensity was scored as 0 (negative), 1 (light brown), 2 (brown), or 3 (dark brown). The percentage of positive cells was scored as 0 (0–5%), 1 (6–25%), 2 (26–50%), 3 (51–75%), or 4 (76–100%). IHC-score = intensity × percentage.
DSP
Continuous sections screened by H&E staining were chosen for spatial transcriptome analysis. The GeoMx WTA panel, which includes 18,677 genes, was utilized to target mRNA expression. Oligonucleotide-tagged WTA panel probes were incubated on the tissue for in situ hybridization. The tissue was then stained using the GeoMx Solid Tumor TME Morphology Kit to differentiate various compartments: PanCK (Cat # GMX-RNA-MORPH-HST-12, Nanostring; green) for epithelial cells, CD68 (sc-20060 AF647, SANTA CRUZ; red) for macrophages, CD66b (ab300122, Abcam; yellow) for neutrophils, and SYTO13 (Cat # GMX-RNA-MORPH-HST-12, Nanostring; blue) for the nucleus. Pathologists distinguished normal epithelial cells from cancerous epithelial cells based on histological morphology. ROIs were selected and evaluated by pathologists. Oligonucleotide tags from the minimum area of illumination or AOI were excised and collected using ultraviolet light irradiation. A total of five types of AOIs were collected: the epithelial cell compartment (PanCK+), macrophage compartment (CD68^+^), neutrophil compartment (CD66b+), stromal compartment (SYTO13+, PanCK-, CD68^−^, CD66b-), and lymphoid structure compartment. According to the manufacturer’s instructions (NanoString), oligonucleotide tags were collected for library preparation, sequencing, and digital quantification on DNBSEQ-T7. In total, 165 AOIs were collected with specific information, which are detailed in Table S1.
Spatial transcriptome data analysis
GeoMx NGS pipeline software V2.0.0.16 was used to convert sorted FASTQ files into DCC files. Subsequently, the DCC file was uploaded to the GeoMx DSP system for further analysis using the data analysis module V2.4.0.421 in the GeoMx DSP control center. The quality control (QC) of transcriptome data included technical signals, technical background, probes, and standardization. When the alignment rate between the reading and the template sequence was <80%, AOI was removed using the technical signal QC. The technical background included three indicators: template free control (TFC) count, negative probe count, and AOI parameters. TFC counting was used to detect template contamination during library construction. AOI with TFC >1000 was removed. In the WTA experiment, negative probe counting was used to measure the overall technical signal level. The threshold for negative probe counting was four counts. AOI parameters were evaluated based on the number of nuclei and surface area. To meet the QC standards, an AOI was required to possess a nuclei count >20 or a surface area >1600 μm2. Meanwhile, we characterized the sequencing saturation for each AOI by the ratio of unique reads to total reads. AOIs with sequencing saturation below 50% were excluded from further analysis. The limit of quantification (LOQ) was calculated for each ROI. LOQ was defined as two standard deviations above the geometric mean of the negative control probe counts, as is standard in NanoString workflow.76 Target values were excluded from analysis if no measurement for that target gene exceeded the LOQ. After filtering of targets using the LOQ, 18346 genes were remained. To ensure consistency across the AOIs, their sizes were standardized by cell number and area normalization. Thereafter, the size of different AOIs was adjusted using 75th percentile normalization to avoid differences between them. Hierarchical clustering and correlation matrix were performed using the “ComplexHeatmap” software package (V2.24.0). Principal component analysis was performed using “prcomp of stats” V4.1.0. For differential expression analysis, edgeR (V4.6.2) was used, with a threshold FC > 1 and p < 0.01 to screen for DEGs. Other related graphs were generated by the “ggplot2″ software package (V3.5.2). For functional annotation and enrichment analysis, divergent genes were processed using the clusterProfiler software package (V4.8.3) with adherence to specific threshold criteria: p value <0.01.
Tumor microenvironment (TME) component estimation
To determine the abundance of immune and stromal cells within the tumor microenvironment (TME), we applied the SpatialDecon (V1.18.0) and ConsensusTME (V0.0.1.9) algorithms to non-epithelial compartments. For SpatialDecon, the expression matrix of non-epithelial compartments were extracted and used as input. The deconvolution algorithm was then employed to calculate immune and stroma scores using default parameters.
For ConsensusTME, the expression matrix of non-epithelial samples was also extracted and used as input. The statMethod parameter was set to “ssgsea”, and the cancer parameter was specified as “ESCA”. The TCGA-ESCA cohort samples were used as a reference for normalization. Within the ssGSEA framework, consensus gene sets were utilized to provide normalized enrichment scores for each cell type, representing their relative abundance across multiple samples.
For subtype-specific analysis of macrophages (MC) and neutrophils (NC), we first leveraged the ESCC single-cell RNA sequencing (scRNA-seq) dataset GEO: [GSE160269](GSE160269) to annotate neutrophil subtypes, identifying two functionally distinct populations: the anti-tumor tumor-associated neutrophil 1 (TN1) and pro-tumor tumor-associated neutrophil 2 (TN2). For macrophage subtypes, we utilized the LM22 signature matrix (integrated in the CIBERSORT algorithm), which includes signatures for M1 (pro-inflammatory, anti-tumor) and M2 (anti-inflammatory, pro-tumor) macrophages, with M2 macrophages referred to as tumor-associated macrophage 2 (TAM2) in the tumor microenvironment context. We then integrated the annotated TN1/TN2 neutrophil signatures with the LM22 matrix to conduct deconvolution analysis via CIBERSORT, excluding normal epithelial samples given the limited biological relevance of tumor-associated subtypes (M1/M2, TN1/TN2) in non-malignant tissues.
Single-cell-RNA-seq data processing
The raw gene expression matrices and relevant clinical data were obtained from GEO: [GSE160269](GSE160269) and performed in R (V4.0.0). Cells expressing less than 300 genes, as well as those with less than 3% ribosomal genes, more than 0.1% hemoglobin genes, and over 20% mitochondrial counts, were removed. Genes expressed in fewer than 3 cells were also excluded. Default parameters of Seurat were utilized, unless otherwise specified. Independent tissue sections’ data were normalized using the SCTransform function of Seurat. Principal component analysis (PCA) was employed for dimensionality reduction and clustering, with a resolution of 0.8 for the first 30 PCAs. Clustering was executed using the FindClusters function, incorporating 30 PCA components and a resolution parameter set to 1.2. Cell types were annotated using known cell-type markers from reference in articles.
Epithelial cells trajectory analysis
To investigate the developmental trajectory of epithelial cells during the progression of ESCC and explore the relationships between specific genes and different stages of ESCC, we applied the SCORPIUS algorithm (V1.0.9), an unsupervised method for inferring linear developmental chronologies from RNA sequencing data. The expression matrix of epithelial compartments was extracted and used as input. We reduced the dataset’s dimensionality using the reduce_dimensionality function with the ‘Euclidean’ parameter. Subsequently, the infer_trajectory function was applied to infer the trajectory of epithelial cells using default parameters. Genes of interest were identified through the gene_importances function with default settings.
Cell-cell communication analysis
The R package CellChat (V1.6.1) was utilized to infer the interplay between cell subpopulations within TME. CellChat objects for advanced stages were created using mESCC and mLN samples, aiming to uncover key ligand-receptor pairs and signaling pathways across different stages and to detect and visualize cell-state-specific cell-cell interactions. Following the official procedure, standardized counts were input into CellChat, followed by standard preprocessing steps. These steps included functions such as identifyOverExpressedGenes, identifyOverExpressedInteractions, and projectData, all executed with default parameter settings. Subsequently, the computeCommunProb, computeCommunProbPathway, and aggregateNet functions were employed to calculate the strength of information flow and communication probability between different cell groups for each ligand-receptor pair. For visualization, methods such as netVisual_bubble and netVisual_aggregate were used to illustrate the interactions and aggregated networks effectively.
mIF staining
The mIF analysis was performed only for good quality samples based on H&E staining of the same specimen. mIF staining was performed on unstained slides of FFPE specimens using the Opal 7-Color Kits (Akoya Biosciences, USA). Six markers were placed in two panels as follows: panel 1 contained pancytokeratin (epithelial marker), CD4 (helper T cell marker), CD8 (cytotoxic T cell marker), CD20 (B cell marker), FOXP3 (regulatory T cell marker), and APOBEC3A; and panel 2 contained pancytokeratin (epithelial marker), CD68 (macrophage marker), CD163 (M2 macrophage marker), CD86 (M1 macrophage marker), CCND1, LAMB1, and IL-18. The stained slides were scanned using Vectra 3.0 multispectral microscope system (Akoya Biosciences, USA) in 10-fold magnification, and the representative ROIs were selected with Phenochart1.0.9 viewer (Akoya Biosciences, USA). The ROIs for mIF analysis were then selected after comparing with H&E slides to capture malignant and premalignant cell clusters and various elements of heterogeneity. The corresponding normal ROIs were selected in the farthest field of tumor periphery with morphological normal tissue on the same slide. Each ROI in panel 1 and 2 was overlapped with sequential sections. The target areas were analyzed using Form 2.4.4 software (Akoya Biosciences, USA). Next, the ROI was divided into two compartments: epithelial compartment and tumor stroma compartment. The individual cells were recognized using 4′,6-diamidino-2-phenylindole (DAPI) nuclei staining and co-localization markers.
Lentiviruses infection
Lentiviruses carrying human OGT cDNA were constructed by Generalbiol (Anhui, China). The full insert sequence was verified by Sanger sequencing. For knockdown, short hairpin RNAs (shRNAs) targeting human OGT were designed and packaged into lentiviral vectors (GenePharma, Suzhou, China). The target sequence for shOGT was 5′-GCACGGCTCTGAAACTTAA-3′. A non-targeting scrambled shRNA was used as the negative control. Lentiviruses were used to infect KYSE150 and KYSE450 cells at a multiplicity of infection (MOI) of 60. Stable cell lines were selected with puromycin (2 μg/mL).
Plasmids and RNA interference
Small interfering RNAs (siRNAs) targeting human OGT and a non-targeting control siRNA were purchased from GenePharma (Shanghai, China). The sequences were as follows.OGT-homo#1 Sense(5′-3′)GCACGGCUCUGAAACUUAATTOGT-homo#1 Antisense(5′-3′)UUAAGUUUCAGAGCCGUGCTTOGT-homo#2 Sense(5′-3′)GAGCAGUAUUCCGAGAAAUTTOGT-homo#2 Antisense(5′-3′)AUUUCUCGGAAUACUGCUCTT
The human OGT coding sequence (GenBank: NM_181672) was amplified and cloned into the BamHI/HindIII sites of the CV702 vector (GeneChem, Shanghai, China). The insert sequence was verified by Sanger sequencing. siRNA and plasmid transfection were performed using GP-transfect-Mate (GeneChem, Shanghai, China) according to the manufacturer’s protocol.
Pharmacological inhibition of cellular functions and PI3K/AKT signaling
To evaluate the activation status of the PI3K/AKT signaling pathway under different OGT or O-GlcNAc conditions, serum starvation followed by serum stimulation was performed. ESCC cells were seeded and cultured under standard conditions until reaching approximately 60–70% confluence. For O-GlcNAc inhibition, cells were treated with the OGT inhibitor OSMI-1 (40 μM) for 40 h. Cells were washed twice with PBS and incubated in serum-free RPMI-1640 medium for 12h to reduce basal PI3K/AKT activity. Where indicated, the PI3K inhibitor LY294002 (10 μM) was added during the final 1h of the starvation period. Subsequently, PI3K/AKT signaling was reactivated by stimulation with complete medium containing 10% fetal bovine serum for 30 min. Cells were immediately lysed in ice-cold RIPA buffer supplemented with protease and phosphatase inhibitors for downstream western blot analysis.
RNA extraction and RT–qPCR
OGT-F:GTCGGCTGCGTGTAGGATATGTGOGT-R:TAGGAAGTTTGTGCCATCGTCTGACTB-F:CACCATTGGCAATGAGCGGTTCACTB-R:AGGTCTTTGCGGATGTCCACGT
Total RNA was extracted from cells and tissue samples using TRIzol reagent (Invitrogen, USA) according to the manufacturer’s instructions. After spectrophotometric quantification of RNA, cDNA was synthesized from RNA using Abm’s 5× All-In-One RT MasterMix, according to the manufacturer’s protocol (Abm, Canada). RT–qPCR was performed using the StepOne Plus RT PCR System (Thermo, USA) with SYBR Green (Vazyme, China). Actin Beta was used as an internal control, and the levels of target genes were calculated based on the 2 −ΔΔCT method. The primers information is as follows.
Cell proliferation and colony formation assays
To assess the effect of OGT on ESCC cell proliferation, cells in the logarithmic growth phase were transfected and, 24 h later, digested with trypsin. After centrifugation, cells were resuspended in complete RPMI-1640 medium to obtain single-cell suspensions and seeded into 96-well plates at a density of 3,000 cells per well in triplicate. Plates were incubated at 37°C with 5% CO_2_. Cell proliferation was measured using the Cell Counting Kit-8 (MedChemExpress, USA). At indicated time points (0–72 h, every 24 h), 10 μL of CCK-8 solution was added to each well, and plates were gently tapped to ensure mixing. After incubation at 37°C for 1–2 h, absorbance was measured at 450 nm using a microplate reader. All experiments were independently repeated three times.
Transwell assays
Transwell assays were used to estimate the migration and invasion of cells. For invasion estimation, Matrigel (Corning, USA) was diluted 1:10 with serum-free RPMI-1640 medium and evenly spread in the chamber at 37°C overnight. Migration assays were performed using Transwell chambers with 8 μm pores (Corning, USA) without Matrigel. Cells were inoculated in the upper chamber at a density of 3 × 10^4^ cells/chamber for 24–48 h. After fixing in paraformaldehyde and staining with crystal violet, three random fields were selected and photographed under an inverted phase contrast microscope (DFC295, Leica, Buffalo Grove, United States), each experiment was independently repeated three times.
Wound healing assay
Cancer cells were seeded into 6-well plates and transfected at ∼70% confluence. After 4–6 h, the medium was replaced, and cells were cultured to ∼90% confluence. A scratch was made using a sterile 200 μL pipette tip guided by a ruler. After washing with PBS, serum-free medium was added. Images were captured at 0 h and 24 h under an inverted microscope. Migration was assessed by measuring wound closure distance and normalizing to the control group. Experiments were performed in triplicate.
Flow cytometry assay
For apoptosis detection, cells (1×10^5^ per test) were washed with cold PBS and resuspended in 100 μL 1× Annexin V Binding Buffer (BD Biosciences, Cat. No. 559763). PE Annexin V (5 μL) and 7-AAD (5 μL) were added, and samples were incubated for 15 min at room temperature in the dark. After adding 400 μL 1× Binding Buffer, stained cells were analyzed by flow cytometry within 1 h.For cell cycle analysis, cells were fixed in 70% ethanol at 4°C overnight, washed with PBS and staining buffer (BD Biosciences, Cat. No. 554656), and resuspended at 1×10^6^ cells per sample. Cells were incubated with 7-AAD (BD Biosciences, Cat. No. 559925) for 30 min at room temperature in the dark and analyzed by flow cytometry. All samples were acquired on a FACS Calibur flow cytometer (BD Biosciences, USA).
Total protein extraction
Forty-eight hours post-transfection, cells were lysed using RIPA buffer supplemented with PMSF (RIPA:PMSF = 100:1), prepared freshly according to cell density. After thorough mixing, the lysis buffer was added to the cells (50 μL per well) and incubated on ice. Cells were vortexed briefly for 10 s every 5 min, for a total of 2–3 cycles, followed by continued incubation on ice for 20–30 min. Lysates were centrifuged at 12,000 rpm for 30 min at 4°C, and the supernatants containing total protein were collected and stored at −80°C for subsequent analyses.
Protein concentration determination
Protein concentrations were quantified using the bicinchoninic acid (BCA) assay. Briefly, a standard curve was generated by adding 0, 1, 2, 4, 8, 12, 16, and 20 μL of BSA standard solution (1 mg/mL) to a 96-well microplate, and adjusting the final volume in each well to 20 μL with PBS. Simultaneously, 1 μL of each protein sample was added to separate wells and brought to 20 μL with PBS.BCA working solution was freshly prepared at a 49:1 ratio of Reagent A to Reagent B, mixed thoroughly, and 200 μL of the working solution was added to each well. Plates were gently mixed and incubated at 37°C for 30 min.After incubation, absorbance was measured at 562 nm using a microplate reader. A standard curve was plotted based on absorbance values of the BSA standards, and sample protein concentrations were calculated accordingly.
Protein denaturation
Measure the protein sample volume precisely and mix with 5× loading buffer at the appropriate ratio. Vortex to mix thoroughly. Heat the mixture in a 100°C metal bath for 5 min to denature proteins. Allow samples to cool to room temperature and store at −20°C.
Western blotting
SDS–PAGE gels were prepared using the following formulation.Component10% Resolving Gel (mL)12% Resolving Gel (mL)5% Stacking Gel (mL)Double-distilled water4.03.33.430% Acrylamide/Bis3.34.00.831.5 M Tris-HCl (pH 8.8)2.52.5–1.0 M Tris-HCl (pH 6.8)––0.6310% SDS0.10.10.0510% APS0.10.10.05TEMED0.0040.0050.005
Gels were freshly prepared prior to electrophoresis. After polymerization, stacking and resolving gels were assembled and used for SDS–PAGE analysis of protein samples.Protein samples (50 μg per lane) were resolved by SDS-PAGE using self-prepared polyacrylamide gels of appropriate concentration. After electrophoresis at a constant voltage of 80 V for 2–3 h, proteins were transferred onto PVDF membranes (pre-activated in methanol) using a wet transfer system at 300 mA constant current on ice. Membranes were blocked in 5% non-fat milk in TBS for 1 h at room temperature, followed by three washes with TBST (10 min each). Membranes were then incubated overnight at 4°C with primary antibodies diluted in TBST. Primary antibodies used were as follows: OGT (1:1000, HUABIO, ET7107-17) O-Linked N-Acetylglucosamine (1:1000, abcam, ab2739), beta Actin (1:1000, Servicebio, GB11001), AKT (1:2500, HUABIO, ET1609-51), p-AKT (1:5000, HUABIO, ET1607-73),PI3K (1:1000, HUABIO, HA601206), p-PI3K (1:1000, Cell signal technology, 4228). The next day, membranes were washed again three times in TBST (10 min each) and incubated with HRP-conjugated secondary antibodies at room temperature for 1 h. After final washes in TBST, target proteins were visualized using enhanced chemiluminescence (ECL) reagents and detected with a digital imaging system. Band intensities were quantified using ImageJ (NIH, USA), with normalization to loading controls as appropriate.
Mass spectrometry analysis
Cells were lysed in 300 μL of lysis buffer, homogenized on ice, and centrifuged at 12,000 × g for 10 min at 4°C. The supernatant was collected, and 50 μg of protein was reduced with 10 mM DTT at 56°C for 1 h, alkylated with 20 mM IAM at room temperature in the dark for 1 h, and quenched with additional DTT. For SP3 digestion, magnetic beads (100 μg/μL) were mixed with samples and 110 μL absolute ethanol, incubated for 15 min, and cleaned with 80% ethanol. Proteins were digested overnight at 37°C with 8 μL trypsin (0.25 μg/μL). Peptides were eluted, freeze-dried, desalted using C18 columns, and dissolved in 0.1% formic acid for MS analysis. LC-MS/MS was performed using a nanoLC system with a 150 μm × 55 mm C18 column. The mobile phases were 0.1% formic acid in water (A) and 80% acetonitrile +0.1% formic acid (B), with a 90 min gradient and 0.4 mL/min flow rate. The gradient of solvent B was: 0 min–8%, 6.5 min–24%, 9 min–36%, 10 min–55%, 11–15 min–99%. Mass spectra were acquired with an MS1 resolution of 240,000 and MS2 resolution of 60,000. The scan range was m/z 380–980, with AGC target 500%, max injection time 5 ms (MS1) and 3 ms (MS2), and collision energy set to 25%. Raw data were analyzed with DIA-NN (v1.9). Search parameters included fixed modification: Carbamidomethyl (C); variable modifications: Oxidation (M), Acetyl (N-term); enzyme: Trypsin; database: Homo_sapiens; max missed cleavages: 2; precursor and fragment mass tolerances: 20 ppm.
Glycoproteome mass spectrometry analysis
Cells were lysed in 300 μL of lysis buffer, homogenized on ice, and centrifuged at 12,000 × g for 10 min at 4°C. The supernatant was collected, and 300 μg of protein was reduced with 10 mM DTT at 56°C for 1 h, alkylated with 20 mM iodoacetamide (IAM) at room temperature in the dark for 1 h, and quenched with an additional 10 mM DTT. Proteins were digested using the FASP method. Samples were transferred to 3 kDa MWCO filters and washed three times with 50 mM TEAB. Then, 28 μL of trypsin (0.25 μg/μL) was added, and digestion was carried out overnight at 37°C. Peptides were collected by centrifugation, rinsed twice with ultrapure water, and lyophilized. For glycopeptide enrichment, dried peptides were incubated with 200 μL of equilibrated HILIC resin in 80% ACN/1% TFA at room temperature for 1 h. The peptide-resin mixture was loaded onto C8 StageTips and washed with 80% ACN/1% TFA. Glycopeptides were eluted using 20% ACN/1% TFA, pooled, and vacuum-dried. Samples were desalted using C18 StageTips and eluted in 80% ACN/0.1% TFA for LC-MS/MS analysis. LC-MS/MS was performed using a PepMap C18 column (150 μm × 170 mm, 1.9 μm, 100 Å) at a flow rate of 600 nL/min over a 131 min gradient. MS acquisition was carried out in DDA mode with MS1 resolution of 60,000, scan range 300–1800 m/z, and MS2 resolution of 15,000. Raw files were analyzed using Byonic (Protein Metrics), with fixed modification: carbamidomethylation (C); variable modifications: oxidation (M), acetylation (N-term); glycan modification: O-glycans (78 human); enzyme: trypsin; database: Homo sapiens (UniProt); peptide and fragment mass tolerance: 20 ppm.
Lung metastasis model
To establish a lung metastasis model, luciferase-labeled KYSE150 cells overexpressing OGT (oeOGT) or vector control were injected into the tail vein of NOD/SCID mice (5 × 10^5^ cells in 100 μL of PBS, n = 5 per group). Bioluminescent imaging was conducted weekly starting from the third week post-injection to monitor metastatic progression. After 7 weeks, the mice were euthanized, and the lungs were harvested and fixed in 4% paraformaldehyde (PFA) at room temperature. Subsequently, gross morphology, hematoxylin and eosin (H&E) staining, and histological quantification of metastatic lesions were performed.
Subcutaneous xenograft model
KYSE450 cells with stable knockdown of OGT (shOGT) or non-targeting control (shNC) were subcutaneously injected into the dorsal flanks of BALB/c-nude mice (5 × 10^6^ cells in 100 μL of PBS/Matrigel [1:1], n = 5 per group). Tumor dimensions were measured every 5 days using a digital caliper, and tumor volume was calculated using the formula: volume = (length × width^2^)/2. After 30 days, the mice were euthanized by CO_2_ inhalation, and tumors were harvested, photographed, and processed for subsequent analysis.
Quantification and statistical analysis
All statistical calculations were performed with the GraphPad Prism 9 software and R (V4.0.0) programs. Dimension reduction analysis was performed using UMAP V.0.2.8.0. Wilcoxon test was used to compare the differences between two groups. Correlation analysis was conducted using Pearson correlation analysis. The significance of overall survival (OS) and progression-free survival (PFS) was analyzed using Kaplan-Meier (K-M) curve analysis based on TCGA-ESCA dataset using the webtool GEPIA2. The K-M curve was compared using the log rank test. The median follow-up time was calculated using the reverse K-M method. The experimental data are presented as the mean value ± standard deviation (SD). Statistical significance was set at p < 0.05.
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1Yang H.Wang F.Hallemeier C.L.Lerut T.Fu J.Oesophageal cancer Lancet 4042024199120053955017410.1016/S 0140-6736(24)02226-8 · doi ↗ · pubmed ↗
- 2Morgan E.Soerjomataram I.Rumgay H.Coleman H.G.Thrift A.P.Vignat J.Laversanne M.Ferlay J.Arnold M.The Global Landscape of Esophageal Squamous Cell Carcinoma and Esophag eal Adenocarcinoma Incidence and Mortality in 2020 and Projections to 2040: New Estimates From GLOBOCAN 2020 Gastroenterology 1632022649658.e 23567180310.1053/j.gastro.2022.05.054 · doi ↗ · pubmed ↗
- 3Wang X.Kutschat A.P.Yamada M.Prokakis E.Böttcher P.Tanaka K.Doki Y.Hamdan F.H.Johnsen S.A.Bromodomain protein BRDT directs ΔNp 63 function and super-enhancer act ivity in a subset of esophageal squamous cell carcinomas Cell Death Differ.282021220722203365870310.1038/s 41418-021-00751-w PMC 8257622 · doi ↗ · pubmed ↗
- 4Li W.Yang Y.Huang L.Yu X.Wang T.Zhang N.Yang M.The TDP-43/TP 63 Positive Feedback Circuit Promotes Esophageal Squamous Cell Carcinoma Progression Adv. Sci.112024 e 240291310.1002/advs.202402913 PMC 1142524839023169 · doi ↗ · pubmed ↗
- 5Shi Y.Yuan Q.Chen Y.Li X.Zhou Y.Zhou H.Peng F.Jiang Y.Qiao Y.Zhao J.Corynoline inhibits esophageal squamous cell carcinoma growth via targ eting Pim-3Phytomedicine 123202415523510.1016/j.phymed.2023.15523538128397 · doi ↗ · pubmed ↗
- 6Chang J.Tan W.Ling Z.Xi R.Shao M.Chen M.Luo Y.Zhao Y.Liu Y.Huang X.Genomic analysis of oesophageal squamous-cell carcinoma identifies alc ohol drinking-related mutation signature and genomic alterations Nat. Commun.820171529010.1038/ncomms 15290 PMC 547751328548104 · doi ↗ · pubmed ↗
- 7Liu X.-Y.Liu Y.B.Xu J.C.Zhang Y.F.Ruan Y.Y.Zhao Y.Wu L.F.Hu J.W.Zhang Z.He M.J.Single-cell transcriptomic analysis deciphers key transitional signatu res associated with oncogenic evolution in human intramucosal oesophag eal squamous cell carcinoma Clin. Transl. Med.132023 e 120310.1002/ctm 2.1203 PMC 997545436855810 · doi ↗ · pubmed ↗
- 8Nagtegaal I.D.Odze R.D.Klimstra D.Paradis V.Rugge M.Schirmacher P.Washington K.M.Carneiro F.Cree I.A.WHO Classification of Tumours Editorial Board The 2019 WHO classification of tumours of the digestive system Histopathology 7620201821883143351510.1111/his.13975 PMC 7003895 · doi ↗ · pubmed ↗