Abstract

   Esophageal squamous cell carcinoma (ESCC) is associated with a highly
   immunosuppressive tumor microenvironment (TME), driven in part by
   cancer-associated fibroblasts (CAFs) that promote immune evasion
   through the secretion of CXCL12. CXCL12 interacts with the CXCR4
   receptor on immune cells, disrupting CD8^+ T cell migration and
   anti-tumor function. To address this, we developed an innovative
   siRNA-based therapeutic approach targeting CXCL12 in CAFs. Using lipid
   nanocarriers (LNCs) as delivery vehicles, we engineered LNCs@si-CXCL12
   nanoparticles to specifically silence CXCL12 expression in CAFs. In
   vitro studies demonstrated that LNCs@si-CXCL12 restored CD8^+ T cell
   migration and inhibited ESCC cell proliferation and migration. In vivo
   experiments in a spontaneous ESCC mouse model showed that CXCL12
   silencing through nanoparticle delivery significantly reduced tumor
   growth, enhanced CD8^+ T cell-mediated tumoricidal activity, and
   improved overall survival. These findings highlight the potential of
   siRNA-loaded nanoparticles targeting CXCL12 as a novel therapeutic
   strategy to reprogram the immunosuppressive TME and enhance immune
   responses in ESCC. This approach provides a promising avenue for
   improving treatment outcomes and overcoming immune evasion in ESCC.

Graphical abstract

   [54]graphic file with name 12951_2025_3476_Figa_HTML.jpg

Supplementary Information

   The online version contains supplementary material available at
   10.1186/s12951-025-03476-x.

   Keywords: Esophageal squamous cell carcinoma, Tumor microenvironment,
   Cancer-associated fibroblasts, CXCL12, SiRNA nanoparticles

Introduction

   Esophageal squamous cell carcinoma (ESCC) is one of the most common
   malignant tumors in the digestive system, involving multiple cell types
   and intricate molecular pathways in its pathogenesis. Research has
   revealed the diverse impacts of genetic factors, environmental factors,
   and lifestyle on ESCC development. For instance, mutations in genes
   such as TP53, CDKN2A, PIK3CA, and MYC lead to cell apoptosis and
   proliferation imbalances, thereby promoting tumor formation
   [[55]1–[56]4]. Additionally, environmental factors like smoking and
   alcohol consumption, along with unhealthy dietary habits and chronic
   gastroesophageal reflux, are significant contributors to the increased
   incidence of ESCC [[57]5–[58]7].

   Cancer-associated fibroblasts (CAFs) play a crucial role in ESCC
   development. They activate the STAT3 signaling pathway by secreting
   cytokines like IL-6, promoting tumor growth and infiltration. They also
   enhance the migration and invasiveness of tumor cells by secreting
   CXCL12, which binds to CXCR4 [[59]8–[60]10]. Furthermore, CXCL12 can
   activate signaling pathways, including PI3K/AKT and MAPK, further
   promoting tumor growth and infiltration [[61]11].

   Tumor fibrosis plays a key role in the occurrence and development of
   cancer, while the application of nanocarriers in gene delivery provides
   new ideas and methods to improve the success rate of tumor treatment.
   Tumor fibrosis refers to the loss of normal structure in the connective
   tissue surrounding tumor cells in the tumor microenvironment (TME),
   resulting in the formation of excessive collagen and other matrix
   components. This process is usually caused by the combined actions of
   cancer cells, CAFs, immune cells, and other components of the TME.
   Fibrosis increases the mechanical strength of the tumor, providing a
   microenvironment that supports tumor cell growth, migration, and
   metastasis. Fibroblasts in the tumor stroma secrete growth factors that
   stimulate tumor cell proliferation. Additionally, tumor fibrosis often
   leads to limited drug uptake and distribution, affecting the efficacy
   of systemic therapies. The density of the tumor stroma may also shield
   against radiotherapy or chemotherapy, thus reducing treatment efficacy.
   Furthermore, the fibrotic TME alters the infiltration pattern of immune
   cells, impairing effective anti-tumor immune responses and allowing the
   tumor to escape immune surveillance. Recent studies have suggested that
   the extent of tumor fibrosis can serve as a prognostic factor, helping
   to predict disease progression and treatment response in patients
   [[62]12–[63]14].

   Nanocarriers, which are micron- or nanometer-sized platforms used for
   drug or gene delivery, can effectively improve therapeutic efficiency.
   In cancer therapy, especially in gene delivery, nanocarriers enhance
   the delivery efficiency of genes by selectively targeting tumor cells
   via surface-modified ligands, while reducing toxicity to normal cells
   [[64]15]. Some gene delivery systems use nanocarriers to improve the
   solubility of water-insoluble gene drugs, thereby enhancing their in
   vivo stability and bioavailability. The design of nanocarriers enables
   controlled release of drugs or genes, extending their action time in
   the TME and improving therapeutic efficacy. Additionally, nanocarriers
   help overcome the biological barriers of the cell membrane, improving
   the efficiency of gene entry into cells, which is critical for most
   gene therapy approaches [[65]16–[66]18].

   The TME plays a crucial role in cancer progression and metastasis by
   influencing various cellular processes such as proliferation, invasion,
   and immune evasion [[67]19]. Chemokines, such as CXCL12 (also known as
   stromal cell-derived factor 1, SDF-1), are key components of the TME
   and regulate interactions between cancer cells and stromal cells. It is
   well-known that CXCL12 promotes tumor growth, angiogenesis, and
   metastasis by binding to its receptor CXCR4, which is often
   overexpressed in various cancers. Therapeutic strategies targeting the
   CXCL12/CXCR4 signaling pathway have become a promising approach in
   cancer treatment [[68]20, [69]21].

   Using small interfering RNA (siRNA) to silence CXCL12 expression is one
   such approach, similar to shRNA (short hairpin RNA). siRNA technology
   can specifically and effectively silence genes by targeting mRNA
   degradation, thereby reducing the expression of the corresponding
   proteins. In the context of the TME, silencing CXCL12 with siRNA can
   disrupt the CXCL12/CXCR4 axis, reducing the signaling through this
   pathway. This disruption can have a profound impact on tumor
   progression, as the CXCL12/CXCR4 signaling pathway is involved in
   processes such as tumor cell migration, invasion, and metastasis. By
   silencing CXCL12 expression in CAFs and other stromal cells, the
   interaction between cancer cells and the microenvironment can be
   altered, potentially reducing tumor invasiveness and metastatic
   potential [[70]22, [71]23]. Studies have shown that siRNA-mediated
   CXCL12 silencing targeting the CXCL12/CXCR4 pathway can inhibit tumor
   growth and metastasis in preclinical models [[72]24, [73]25].
   Additionally, this method can enhance the efficacy of other anti-cancer
   therapies by making tumor cells more sensitive to treatment. However,
   further studies are needed to comprehensively understand the impact of
   CXCL12 silencing on the TME and optimize the use of siRNA therapy in
   cancer treatment.

   Nanotechnology targeting CAFs provides an innovative therapeutic
   strategy, precisely delivering therapeutic drugs to key TME cell types
   [[74]26]. For example, a tailored PLGA nanoparticle formulation can
   simultaneously deliver chemotherapy drugs and siRNA targeting HGF,
   effectively transforming activated CAFs into a quiescent state, thus
   significantly inhibiting tumor proliferation, migration, and invasion
   and improving tumor permeability [[75]27]. Additionally, Bingyuan Fei
   and colleagues investigated a novel nanomedicine that modulates CAFs,
   normalizes tumor vasculature, and sequentially sensitizes photodynamic
   therapy (PDT), effectively enhancing drug accumulation in the tumor,
   alleviating hypoxia in the TME, and ultimately boosting the immune
   response induced by PDT [[76]28].

   Therefore, based on the above background, this research project will
   comprehensively analyze the interactions between CAFs and ESCC cells
   using transcriptome and single-cell RNA-seq technologies. The project
   aims to design nano-spheres targeting CAFs to deliver si-CXCL12,
   silencing CXCL12 gene expression to modulate the TME and enhance the
   anti-tumor response of CD8^+ T cells. Through this highly precise
   nano-delivery system, we seek to uncover new mechanisms by which CAFs
   regulate tumor growth and immune evasion in ESCC and provide potential
   new strategies for clinical therapy.

Materials and methods

Clinical sample collection

   This study adhered to the ethical principles of the Helsinki
   Declaration and was approved by our hospital's Clinical Ethics
   Committee (No. 2022-197-01). Written informed consent was obtained from
   all registered patients. Ten ESCC patients without prior drug treatment
   or other underlying diseases were selected at our institution for the
   surgical acquisition of cancer tissue and adjacent normal tissue
   [[77]29, [78]30]. The ESCC cancer tissue and corresponding adjacent
   normal tissue were fixed in formalin and embedded in paraffin for
   histopathological diagnosis. All tissue sections were evaluated by
   experienced pathologists and patients'TNM staging was classified
   according to the seventh edition TNM staging standard published by the
   AJCC [[79]31]. The clinical and pathological characteristics of ESCC
   patients are presented in Table S1. Regular follow-ups were conducted
   for all patients to obtain their clinical outcomes.

Bulk RNA-seq analysis

   Eighty-two ESCC bulk RNA-seq data were downloaded from The Cancer
   Genome Atlas (TCGA) database ([80]https://portal.gdc.cancer.gov/). The
   tumor cell composition in the tissue was analyzed using the R language
   packages “CIBERSORT” (immunocyte proportions), “MCPcounter” (cell
   abundance scores), and “xCell” (cell enrichment scores). Samples were
   grouped into high and low Fibroblasts based on the median calculation
   results for subsequent gene differential analysis, taking the
   intersection [[81]32].

10 × genomics

   Fresh tumor tissues and adjacent non-tumor tissues from three ESCC
   patients were collected for single-cell RNA sequencing (scRNA-seq). The
   samples were washed with 0.04% PBS-BSA, minced with a surgical blade,
   and digested with papain enzyme (P4762, Sigma, Germany) at 37 °C for 30
   min. Following digestion, the cell pellets were resuspended in a
   DNase-inhibitor solution. The suspended single cells were filtered
   through a pre-wetted 40 μm cell strainer with HBSS, and then
   centrifuged at 300×g for 5 min at 4 °C. Subsequently, the cell pellets
   were resuspended in 10 ml of 30% Percoll in PBS, centrifuged at 700×g
   for 10 min. After removing the surface debris, the cells were
   resuspended in FACS buffer (PBS, 1% BSA) and stained with Zombie NIR
   Fixable Viability Kit (423105, BioLegend, USA) for viability
   assessment. FACS sorting was performed using BD FACS Aria II cell
   sorter, and the sorted cells were collected in a capture medium (PBS
   containing 0.04% BSA). Cell counting was done using a CountessTMII
   automated cell counter with trypan blue staining. Cells with higher
   than 80% viability were loaded onto the 10 × Genomics Chromium chip
   following the manufacturer’s instructions [[82]33].

Single-cell cDNA library preparation and quality control analysis of
scRNA-seq data sets

   Gel bead-in-emulsion (GEM) was generated by mixing single-cell
   suspension, gel beads, and oil through the 10 × Genomics chromium
   controller. After droplet formation, the sample was added to PCR tubes
   and subjected to reverse transcription using the T100 Thermal Cycler
   (Bio-Rad): incubation at 53 °C for 45 min, followed by 85 °C for 5 min,
   and then held at 4 °C. The cDNA was synthesized, amplified, and
   evaluated for quality using the Agilent Bioanalyzer 2100. For library
   construction, P5 primers, Read 2 (reads 2 sequencing primer sites),
   Sample Index, and P7 primers were added. The resulting library
   underwent quality control before sequencing using Illumina HiSeq4000
   PE125.

   Further analysis of scRNA-seq data was performed using 10 × Cell Ranger
   (version 2.2.0). Illumina sequencing machine base calling files (BCLs)
   were converted to generate FASTQ files, followed by alignment,
   filtering, barcode, and UMI counting. The small mouse reference
   transcriptome (mm10) was aligned using STAR (Spliced Transcripts
   Alignment to a Reference). Cell Ranger conducted primary quality
   control (QC) on the FASTQ files to produce high-quality data [[83]33].

Single-cell transcriptomic analysis

   The “Seurat” package in R software was utilized to analyze single-cell
   data, following quality control standards of nFeature_RNA > 50 &
   percent.mt (5. To reduce the dimensionality of the scRNA-Seq data set,
   principal component analysis (PCA) based on the top 1500 highly
   variable genes was performed. The Elbowplot function provided by the
   Seurat package was used to select the top 20 principal components (PCs)
   for downstream analysis. Clustering identification was conducted using
   the FindClusters function by Seurat, with a resolution set to the
   default value (res = 1). Subsequently, the t-SNE algorithm was applied
   to reduce nonlinear dimensionality of the scRNA-seq sequencing data.
   Seurat package was used to filter marker genes for various cell
   subgroups. Lastly, cell annotation was performed based on known cell
   lineage-specific marker genes, and cell communication analysis was
   carried out using the “CellChat” package [[84]34, [85]35].

Isolation and cultivation of CAFs and NFs

   ESCC-related CAFs and paired normal fibroblasts (NFs) were isolated
   from fresh tumor tissue and adjacent non-tumor tissue (at least 5 cm
   away from the tumor edge) of three ESCC patients. Tissues were cut into
   small pieces (1–2 mm^3), washed with PBS, digested with 1 mg/mL type II
   collagenase (1,148,090, Sigma-Aldrich, USA) at 37 °C in a 5% CO[2]
   environment for 2 h, followed by filtration and centrifugation to
   collect cell pellets. The cell pellets were resuspended in RPMI 1640
   medium (12,633,012, ThermoFisher, USA) containing 10% FBS (10,099,141,
   ThermoFisher, USA), plated in 25 cm^2 culture flasks, and after 30 min,
   non-adherent cells (primarily tumor cells, as fibroblast adhesion time
   (< 30 min) is much shorter than tumor cells adhesion time (typically
   over 1 h)) were removed. Fibroblasts were identified after 2–3 passages
   using immunofluorescence for fibronectin and Cytokeratin to distinguish
   primary NF/CAF isolates (Figure S1A). Subsequently, CAF marker α-SMA
   and FAPα were identified using immunofluorescence (Figure S1B),
   resulting in uniform NFs or CAFs cells for further analysis [[86]36].

Isolation of peripheral blood mononuclear cells (PBMC) from human peripheral
blood

   Blood samples were collected from 10 healthy voluntary blood donors,
   all of whom provided written informed consent and complied with the
   ethical principles of the Helsinki Declaration. The study was approved
   by the Clinical Ethics Committee of our institution (No. 2022-197-01).
   The PBMCs were isolated using a PBMC isolation kit (P8610, Beijing
   Solebao Technology Co., Ltd., China). Fresh anticoagulated whole blood
   supplemented with 20 U/mL heparin (H8060, Beijing Solebao Technology
   Co., Ltd., China) was diluted with an equal volume of PBS. The diluted
   blood was layered onto the separation solution in a centrifuge tube,
   ensuring a clear interface between the two liquids. The tube was
   centrifuged at room temperature at 1000×g for 30 min, resulting in
   visible layering. The white membrane layer containing cells was
   carefully collected and transferred to a clean 15 mL centrifuge tube.
   The cells were washed with 10 mL PBS, centrifuged at 250×g for 10 min,
   the supernatant was discarded, and the cells were resuspended in 5 mL
   PBS and centrifuged at 250×g for another 10 min. The supernatant was
   removed, and the cells were resuspended for further use.

   CD8 T cells were isolated from PBMC using CD8 MicroBeads (130-045-201,
   Miltenyi Biotec, Germany). The purity of CD8 T cells was confirmed to
   be 95% using anti-CD3 cell antibody (981,004, Biolegend, USA), with the
   Jurkat cell line (iCell, h117) serving as a control. Jurkat is a
   well-known cell line with a high frequency of CD88^+ T cells, and
   anti-CD8 cell antibody (344,722, Biolegend, USA) with a flow cytometer
   (Figure S2) [[87]37]. All experiments were carried out with CD8 T cells
   at a concentration of 1.5 × 10^6 cells/mL in X-VIVO 15 medium (04-418Q,
   LONZA, Switzerland) containing 10% FBS. The cells were stimulated with
   100 IU/mL IL-2 (130-097-743, Miltenyi Biotec, Germany) and 10 μL/mL
   CD3/CD28 T cell activator (11131D, Thermo Fisher Scientific, USA) for
   48 h [[88]38].

Cell culture and lentiviral transduction

   The human ESCC cell line TE-1 was purchased from Shanghai iCell
   (iCell-h211, Shanghai, China), and Het-1A was obtained from ATCC
   (CRL-2692). The cells were cultured in RPMI 1640 medium (11,875,119,
   Thermo Fisher, USA) supplemented with 10% fetal bovine serum and 1%
   penicillin–streptomycin (15,140,148, Gibco, CA, USA), and maintained in
   a humidified atmosphere with 5% CO2 at 37°C. Cell passaging was
   performed when the cell growth density reached approximately 80%.

   For the construction of lentiviral plasmids, the overexpression plasmid
   pHAGE-CMV-MCS-IRES-ZsGreen, helper plasmids pSPAX2 and pMD2.G, shRNA
   plasmid pSuper-retro-puro, and helper plasmids gag/pol, VSVG were
   co-transfected into 293 T cells (CL-0005, Wuhan PuNuosi Life Technology
   Co., Ltd., Hubei, China) using Lipo2000 (11,668,500, Invitrogen, USA).
   The cell culture supernatant was collected after 48 h, filtered through
   a 0.45 μm filter, and the virus was harvested by centrifugation. The
   concentrated virus was collected after 72 h, and the titers were
   determined by mixing the two batches of virus.

   For stable cell line construction, cells in the logarithmic growth
   phase were digested with trypsin, resuspended to a concentration of 5 ×
   10^4 cells/mL, and seeded into 6-well plates at 2 mL per well. After
   overnight incubation at 37 °C, cells were infected with lentivirus at a
   concentration of 100 μL/mL. After 48 h, the efficiency of GFP
   expression was observed under a fluorescence microscope. Stable cell
   lines were established by changing the culture medium to a complete
   medium containing 2 μg/mL puromycin (A1113803, Gibco, USA) 72 h
   post-infection and continued for 5 days. The expression levels of
   target genes in different groups of cells were then detected by Reverse
   Transcription Quantitative Polymerase Chain Reaction (RT-qPCR).

   CAF cell grouping: Control (no treatment group), sh-NC (sham lentivirus
   control group), sh-CXCL12-1 (sh-CXCL12 knockdown lentivirus group 1),
   sh-CXCL12-2 (sh-CXCL12 knockdown lentivirus group 2); CD8 T cell
   grouping: sh-NC (sham lentivirus control group), TE-1 + NFs (co-culture
   of TE-1 and NFs group), TE-1 + CAFs (co-culture of TE-1 and CAFs
   group), TE-1 + CAFs-sh-NC (co-culture of TE-1 and CAFs-sh-NC group),
   TE-1 + CAFs-sh-CXCL12 (co-culture of TE-1 and CAFs-sh-CXCL12 group);
   CAFs-sh-NC (CD8^+T cell and CAFs-sh-NC co-culture group),
   CAFs-sh-CXCL12 (CD8^+T cell and CAFs-sh-CXCL12 co-culture group). The
   lentivirus was purchased from Shanghai Hanheng Biotechnology Co., Ltd.
   (Shanghai, China), and this company completed the primer sequences and
   plasmid construction. All experimental procedures were performed
   according to the manufacturer’s instructions and repeated three times.
   The shRNA sequences can be found in Table S2 [[89]39].

Co-culture system of CD8^+T cells and CAFs

   Isolated CD8^+T cells were cultured and expanded under CD3/CD28/IL2
   stimulation, while CAFs were transduced with sh-CXCL12 lentivirus.
   Twelve hours before co-culture with CD8 T cells, CAFs were seeded in a
   culture plate at a density ten times that of CD8^+T cells. After 48 h
   of co-culture, cells were fixed for immunofluorescence staining, and
   CD8^+T cells were collected for flow cytometry analysis [[90]40].

Co-culture system of CAFs and TE-1 cells

   To evaluate the impact of NFs and CAFs on TE-1 cells, a Transwell
   co-culture system was utilized. NFs or CAFs were co-cultured with TE-1
   cells using 0.4 µm Transwell chambers (3412, Corning, USA). The cells
   were adjusted to a density of 1 × 10^5/mL, with TE-1 cells seeded in
   the upper chamber and 1.5 mL of CAFs or NFs in the lower chamber.
   Co-cultures were maintained for 48 h, following which TE-1 cells were
   extracted for further analysis [[91]41].

RT-qPCR

   Total RNA from tissues or cells was extracted using TRIZOL (15,596,018,
   Invitrogen, USA) following the manufacturer’s instructions, and RNA
   concentration was determined. For mRNA analysis, cDNA was synthesized
   using the reverse transcription kit (4,368,814, Applied Biosystems,
   USA) per the manufacturer’s protocol, with the cDNA diluted to 50 ng/μL
   for subsequent quantitative PCR. Each reaction consisted of 2 μL of
   cDNA in a total volume of 25 μL, and amplification was done using the
   ViiA 7 Real-Time PCR System (Applied Biosystems, USA). Primers used in
   this study were synthesized by the Dalian Takara Company (Table S3).
   Relative gene expression levels were calculated using the comparative
   cycle threshold method (2^−∆∆CT method) with β-actin as the internal
   control [[92]42], where ∆∆Ct = ∆Ct [test]–∆Ct [control], ∆Ct = Ct
   [target]–Ct [internal], and relative gene expression level = 2^−∆∆Ct.
   Each experiment was repeated three times.

Enzyme-linked immunosorbent assay (ELISA)

   The concentrations of relevant factors in cell culture supernatants
   were determined using human CXCL12 ELISA kit (JL19499, Shanghai
   Jianglai Biological, China), human CXCL11 ELISA kit (ab289695, Abcam,
   UK), human CXCL9 ELISA kit (JL14160, Shanghai Jianglai Biological,
   China), human CXCL10 ELISA kit (JL11028, Shanghai Jianglai Biological,
   China), and human CCL4L2 ELISA kit (LS-C521156, LifeSpan Biosciences,
   USA) following the respective kit instructions [[93]43].

Western blot

   Cell pellets were lysed in 200 μL of Lysis Buffer in 1.5 mL Eppendorf
   tubes. After homogenization, the tubes were placed on ice for 10 min,
   centrifuged (3000×g, 4 ℃, 10 min), and the supernatant was transferred
   to pre-chilled Eppendorf tubes. The protein concentration was measured
   using the BCA assay, and the samples were denatured in Loading Buffer
   (99 ℃, 10 min) and stored at − 20 ℃. Proteins were separated based on
   molecular weight, with 20 μL of each sample loaded per well. PVDF
   membranes were activated with methanol and proteins were transferred
   using a BioRad electrophoresis system with adjusted voltage or current
   based on protein size. The membranes were blocked with 5% BSA at room
   temperature for 1 h and then incubated with respective primary
   antibodies against CXCL12 (ab155090, 1:5000, Abcam, UK), CXCR4
   (ab181020, 1:1000, Abcam, UK), GAPDH (ab8245, 1:1000, Abcam, UK) at 4 ℃
   overnight. The membranes were washed with TBST (containing 0.1% Tween
   20) and incubated with secondary antibodies at room temperature for 1
   h, followed by additional washes. The chemiluminescent substrate was
   applied, images were captured using a gel imaging system, and protein
   bands were quantified using AlphaView SA software (Version: 3.4.0)
   [[94]44].

CCK-8 experiment

   The proliferation of cells was assessed using a CCK-8 assay kit
   (40203ES60, Yeasen, Shanghai, China). Log-phase cells were taken and
   adjusted to a concentration of 5 × 10^4 cells/mL with a complete
   culture medium. The cells were seeded into a 96-well culture plate,
   with 100 μL of cell culture medium added to each well. The plate was
   then incubated in a cell culture incubator for 0 h, 24 h, 48 h, and 72
   h. After discarding the supernatant, the fresh culture medium was
   added, followed by 10 μL of CCK-8 solution to each well. The plate was
   then further incubated at 37 °C for 2 h. The absorbance (A) was
   measured using a Multiskan FC microplate reader (51,119,080, Thermo
   Fisher Scientific, USA) at a wavelength of 450 nm. The proliferation
   rate (%) was calculated as [(A [control group]–A [experimental
   group])/A [control group]] × 100% [[95]45, [96]46]. Three replicate
   wells were set up for each group, and the experiment was repeated three
   times.

EDU staining

   Cell proliferation was detected using an EDU assay kit (C0075S,
   purchased from Biyuntian, Shanghai, China). Cells were treated with
   EdU(10 μM)-containing medium at different conditions, then incubated at
   37 °C for 2 h. The culture medium was removed, and the cells were fixed
   with 1 mL of 4% paraformaldehyde at room temperature for 15 min. After
   washing with PBS three times for 3 min each, the cells were
   permeabilized with 1 mL of 0.3% Triton X-100 in PBS at room temperature
   for 15 min. The cells were then stained with Hoechst 33,342 for 10 min,
   mounted, observed and photographed under a microscope (IX73, OLYMPUS,
   Japan). Five random fields were captured for each sample for counting,
   and the rate of EDU-positive cells was calculated [[97]47]. The
   experiment was repeated three times.

CD8^+ T cell cytotoxicity and exhaustion detection

   CD8 T cells were treated differently, cultured for 4 days, and then
   collected. The cells were resuspended in 2% FBS in PBS and labeled with
   PE/Cy7-conjugated Anti-CD8 (ab217344, 1:500, Abcam, UK), Anti-Granzyme
   B (ab255598, 1:200, Abcam, UK), Anti-perforin (ab47226, 1:1000, Abcam,
   UK), Anti-PD-1 (ab300425, 1:1000, Abcam, UK), and Anti-IFN-gamma
   (ab224197, 1:1000, Abcam, UK) at 4 °C for 30 min. The cells were washed
   twice with 2% FBS in PBS and analyzed using a BD LSRFortessa flow
   cytometer (BD Bioscience, New Jersey, USA). Data analysis and
   quantification of experimental results were performed using BD FACSDiva
   software [[98]48–[99]51].

Flow cytometry analysis of cell apoptosis

   Apoptosis of cells was assessed using the Annexin V-FITC/PI double
   staining method. Cells were seeded at a density of 2 × 10^5 cells/well
   in 6-well plates. The cells were collected in a 15 mL centrifuge tube,
   centrifuged at 800 g, and the supernatant was discarded. The cell
   pellet was washed twice with PBS and resuspended in 500 μL binding
   buffer according to the instructions of the BD Bioscience cell
   apoptosis assay kit (556,547, BD Bioscience, USA). Subsequently, 5 μL
   of FITC and 5 μL of PI were added to the suspension, mixed well, and
   incubated for 15 min. Cell apoptosis was detected using a BD
   FACSCalibur flow cytometer [[100]52]. Annexin V-FITC positive cells
   were considered apoptotic cells. The experiment was repeated three
   times.

Transwell assay for invasion and migration

   For invasion and migration assays, 8 μm Transwell inserts (3422,
   Corning, USA) in 24-well plates were used. For invasion assay, matrix
   gel (no need for migration assay) was added to each insert and
   incubated at 37 °C for 2 h before the experiment. ESCC cells were
   digested, washed with PBS, resuspended in serum-free culture medium,
   and adjusted to a cell density of 3 × 10^5 cells/mL. Three inserts were
   prepared for each group, with 200 μL of cell suspension added. The
   lower chamber was filled with 700 μL of complete culture medium, and
   the inserts were incubated at 37 °C with 5% CO[2] for 48 h. After
   fixation with methanol for 30 min, the inserts were stained with 0.05%
   crystal violet for 5 min, and photographed under a microscope (IX73,
   OLYMPUS, Japan) after removing internal cells with a cotton swab.
   Migration assay followed the same steps as invasion assay but without
   adding matrix gel. Images were processed and quantified using ImageJ
   software [[101]29]. The experiment was repeated three times.

Immunofluorescence staining

   Cells were seeded in confocal cell dishes. After 24 h, the cells were
   washed twice with PBS, or tissue slices were fixed in 4% PFA for 15
   min, washed three times with 0.3% PBST for 5 min each time. The cells
   were then blocked with 5% goat serum (diluted in 0.3% PBST) at room
   temperature for 1 h. Subsequently, the cells were incubated with
   diluted primary antibodies including rabbit anti- α-SMA (1:50,
   14,395-1-AP, Proteintech, USA), mouse anti-FAP (1:50, sc-65398, Santa
   Cruz Biotechnology, USA), rabbit anti-CD8 (ab237709, 1:50, Abcam, UK),
   and mouse anti-GZMB (MA1-80734, 1:100, Invitrogen, USA) at room
   temperature for 1 h. The cells were washed three times with 0.3% PBST
   for 10 min each time. Then, the cells were incubated with secondary
   antibodies conjugated with red fluorescent protein (Cy3) including goat
   anti-rabbit secondary antibody (ab6939, 1:500, Abcam, UK), or with
   secondary antibodies conjugated with green fluorescent protein (FITC)
   including goat anti-mouse antibody (ab6785, 1:500, Abcam, UK), or with
   secondary antibodies conjugated with green fluorescent protein (FITC)
   including goat anti-rabbit secondary antibody (ab6717, 1:500, Abcam,
   UK) at room temperature for 1 h. After washing three times with 0.3%
   PBST for 10 min each time, the coverslips were air-dried, and the cells
   were observed and photographed under a microscope (IX73, OLYMPUS,
   Japan) [[102]44, [103]53].

3D spheroid invasion assay: in this assay

   3D MCTS developed for TE-1 cells were embedded in basement membrane
   extract (R&D Systems, MN, USA) to provide extracellular matrix around
   the tumor spheroids. The embedded spheroids were treated with ARV, VEM,
   and their combination, each at a concentration of 1 μM. The treatment
   protocol followed was similar to the 3D spheroid cell viability assay,
   during which tumor growth, invasion percentage, and invasive pseudopod
   length were analyzed. Images of the MCTS were captured every other day
   using the Evos imaging system (Thermo Fisher Scientific, MA, USA)
   [[104]54].

Serum stability test

   Nanoparticles were incubated with fresh human serum at a concentration
   of 0.5 nmol/mL for 24 h. Protein binding in human serum was measured
   using Sephadex G-50 size exclusion chromatography (GE Healthcare
   Illustra, UK). 25 µL of serum samples were tested at different time
   points.

Subcutaneous transplantation and metastasis models in humanized CD34^+ mouse
models

   Sixty CD34^+ humanized mice, huHSC-(M-NSG) (NM-NSG-017), were purchased
   from Shanghai Model Organisms Center, Inc. (Shanghai, China) and housed
   in SPF-grade animal facilities with a humidity of 60–65% and
   temperature of 22–25 °C. After acclimating for one week, the mice were
   observed for their health status before the experiment.

   In the subcutaneous transplantation experiment, CD34^+ humanized mice,
   huHSC-(M-NSG), were injected subcutaneously with TE-1 cells (5 mice per
   group). TE-1 cells (4 × 10^6 cells/mouse) were mixed with CAFs (2
   × 10^5 cells/mouse), suspended in a mixture of PBS and Matrigel (1:1),
   and then injected subcutaneously into the right axilla of the mice.
   Tumor length (L), width (W), and volume (V) were measured every other
   day starting from day 7. Tumor volume was calculated using the formula
   V = (L × W^2) × 0.5, and a tumor growth curve was plotted. After 35
   days, the mice were euthanized with 150 mg/kg isoflurane
   (Sigma-Aldrich, USA), and tumors were excised, weighed, and
   photographed. The subcutaneous transplantation experiment groups
   included: huHSC-(M-NSG) mice injected with TE-1 + NFs (co-injection of
   TE-1 and NFs cells), TE-1 + CAFs (co-injection of TE-1 and CAFs cells),
   TE-1 + CAFs-sh-NC (co-injection of TE-1 and CAFs-sh-NC cells), and TE-1
   + CAFs-sh-CXCL12 (co-injection of TE-1 and CAFs-sh-CXCL12 cells)
   [[105]44].

   In the lung metastasis model experiment (6 mice per group), TE-1 and
   CAFs mixed cells were co-cultured for 48 h, TE-1 stable cells were
   digested with trypsin to obtain single-cell suspension, and 10^6 TE-1
   cells (in 100 μL PBS) were injected into the mice via tail vein. After
   5 weeks, mice were euthanized with 150 mg/kg isoflurane, lungs were
   isolated, and lung metastases were observed through H&E staining and
   histological examination. This experiment protocol and animal use were
   approved by the Institutional Animal Care and Use Committee (No.
   GZRCH-AE-20211012). The lung metastasis model experiment groups
   included: huHSC-(M-NSG) mice injected with TE-1 + NFs (TE-1 cells
   co-cultured with NFs before injection), TE-1 + CAFs (TE-1 cells
   co-cultured with CAFs before injection), TE-1 + CAFs-sh-NC (TE-1 cells
   co-cultured with CAFs-sh-NC before injection), and TE-1
   + CAFs-sh-CXCL12 (TE-1 cells co-cultured with CAFs-sh-CXCL12 before
   injection) [[106]55].

Preparation and characterization of lipid nanocarriers (LNCs)@si-CXCL12

   Materials and reagents required for the preparation and
   characterization of LNCs@si-CXCL12 include
   1,2-distearoyl-sn-glycero-3-phosphoethanolamine-N-amino-2000
   (DSPE-PEG2000) and DSPE-PEG2000-maleimide (DSPE-PEG-Mal) purchased from
   Avanti Polar Lipids; stearic acid (SA) and polyethyleneimine (PEI 1800)
   purchased from Aladin; 1-[3-(dimethylamino)propyl]−3-ethylcarbodiimide
   hydrochloride (EDC) and N-Hydroxysuccinimide (NHS) from J&K Scientific.
   The si-CXCL12 (5’-CCAGAGCCAACGTCAAGCATCTGAA-3’) was synthesized by
   GenePharma.

   Firstly, SA was dissolved in DMSO, EDC and NHS were added, and the
   mixture was stirred at room temperature for 2 h to activate the
   carboxyl groups on SA. Subsequently, PEI 1800 was added, and the N/P
   ratio was set to 15:7 [[107]56] and the stirring continued for 24 h.
   The residual SA was extracted using ethyl acetate, then further
   purification of the obtained PSA through dialysis and freeze-drying. By
   performing the Michael addition reaction, Anti-FAPα (PA5-99458,
   Invitrogen, USA) and DSPE-PEG2000-Mal were coupled in a
   chloroform/methanol solution, and the final product was obtained by
   dialysis and stored at – 20 °C. DSPE-PEG, DSPE-PEG-Anti-FAPα, and PSA
   were dissolved in methanol at a mass ratio 8:2:1 to form a liposomal
   membrane, which was hydrated with PBS and sonicated to form micelles.
   Finally, si-CXCL12 was mixed with the micelles at a weight ratio of
   1:5, and the mixture was incubated at room temperature for 15 min to
   form LNCs@si-CXCL12 complex [[108]57]. Dynamic light scattering (DLS)
   and Zeta potential analysis were used to determine the particle size
   and surface charge of the self-assembled complex, and the morphology of
   the complex was observed under transmission electron microscopy (TEM).

Stability test of LNCs@si-CXCL12

   The physical stability of freshly prepared LNCs@si-CXCL12 was monitored
   at different time points, including particle size, zeta potential, and
   encapsulation efficiency (%), for a period of 3 months. During this
   period, LNCs@si-CXCL12 were stored at 4 °C [[109]58].

Establishment of spontaneous ESCC mouse model

   Beijing Vital River Laboratory Animal Technology Co., Ltd. (strain
   code: 219) provided female C57BL/6 mice aged 6 to 7 weeks. The mice
   were housed in specific pathogen-free (SPF) animal facilities with
   controlled room temperature and humidity at 40–60%, following a 12 h
   light–dark cycle.

   To induce spontaneous ESCC, mice were fed with drinking water
   containing 100 μg/mL 4-nitroquinoline 1-oxide (4-NQO, HY-33354, MCE,
   USA) continuously for 16 weeks. Subsequently, mice were provided with
   sterile pure water to promote ESCC formation. Throughout the
   experiment, the mice's activity, water intake behavior, body weight,
   and survival status were observed and recorded every two days. At weeks
   16, 30, and 32 of feeding, the entire esophageal tissue of the mice was
   stained with H&E to detect the induced ESCC process. Additionally,
   total RNA was extracted from the esophageal tissue of mice fed for 28
   weeks to detect the expression of CXCL12 and CXCR4. Ki67 and α-SMA
   staining were also performed on the esophageal tissue. At week 30 of
   the experiment, 4-NQO-induced ESCC mice were randomly divided into two
   groups and intraperitoneally injected with 200μL PBS or 10mg/kg
   LNCs@si-CXCL12 every 2 days for two weeks. After the treatment, the
   esophageal tissues of the mice were collected, photographed using a
   dissecting microscope, and the number and diameter of tumor lesions
   were recorded [[110]59].

H&E staining

   H&E staining was performed using the hematoxylin and eosin staining kit
   (C0105, Beyotime, China). Briefly, lung tissues were fixed in 10%
   neutral buffered formalin at 4 °C for 24 h, followed by dehydration,
   embedding in paraffin, sectioning, deparaffinization in xylene,
   gradient alcohol hydration, and distilled water rinsing. The tissues
   were then stained with hematoxylin for 5–10 min; excess stain was
   removed with deionized water for approximately 10 min, followed by
   eosin staining for 30 to 2 min. Subsequently, the tissues were
   dehydrated in gradient alcohol, cleared in xylene [[111]29, [112]30],
   and mounted using neutral resin or other mounting medium. Slides were
   observed and photographed under an inverted microscope (IX73, OLYMPUS,
   Japan).

Immunohistochemistry staining

   Tissue blocks were fixed in 4% paraformaldehyde phosphate buffer
   solution for 12 h. Following routine deparaffinization in xylene and
   gradient alcohol hydration, the tissues were boiled in 0.01M citrate
   buffer for 15–20 min, cooled to room temperature, PBS washed, and
   incubated in freshly prepared 3% H[2]O[2] methanol for 10 min at room
   temperature. Subsequently, the tissues were incubated with 10% goat
   serum blocking solution (16210072, Gibco, USA) for 20 min at room
   temperature. Primary antibodies against CXCL12 (17402-1-AP, 1:100,
   Proteintech, USA), CXCR4 (ab181020, 1:100, Abcam, UK), and α-SMA
   (ab5694, 1:200, Abcam, UK) were added and incubated for 1 h at room
   temperature, followed by PBS wash. Secondary antibodies, goat
   anti-rabbit IgG (ab6721, 1:500, Abcam, UK) or goat anti-mouse IgG
   (ab6789, 1:500, Abcam, UK), were then added and incubated for 1 h at
   room temperature. After PBS wash, the samples were incubated with SP
   (streptavidin-peroxidase) for 30 min at 37 °C, followed by DAB color
   development for 5–10 min, stop reaction with water for 10 min,
   counterstained with hematoxylin for 2 min, differentiated in
   hydrochloric acid ethanol, water rinsed for 10 min, dehydrated,
   cleared, and mounted. The stained slides were observed and photographed
   under a microscope (IX73, OLYMPUS, Japan), processed and quantified
   using ImageJ. Staining intensity and extent of positive staining were
   scored, and the scores were summed for comparison [[113]30]. Three
   pathologists independently performed slide scoring, and agreement
   between observers was achieved. Staining intensity was graded as
   follows: 0 (negative), 1 (weakly positive), 2 (moderately positive),
   and 3 (strongly positive). The percentage of positive cells was also
   divided into four categories: 0–25% positive cells scored as 1; 26–50%
   as 2; 51–75% as 3; and more than 75% as 4. A total score of ≤ 1 defined
   negative and a score > 1 defined positive.

Statistical analysis

   The research data was analyzed using SPSS software (version 21.0, IBM,
   USA). Continuous data were presented as mean ± standard deviation.
   Normality and homogeneity of variance were initially assessed. If the
   data met the criteria of normal distribution and homogeneity of
   variance, non-paired t tests were applied for intergroup comparisons,
   while one-way analysis of variance or repeated measures analysis of
   variance were used for comparisons among multiple groups. Pearson
   correlation analysis was utilized to examine the relationship between
   two variables. A significance level of P < 0.05 was considered
   statistically significant.

Results

Bioinformatics analysis reveals close association of chemokines (CXCL12,
CXCL1, CXCL8, CCL20) with ESCC-CAFs

   The TME is mainly composed of immune cells, CAFs, endothelial cells,
   and extracellular matrix, playing crucial roles in promoting tumor cell
   proliferation, inhibiting tumor cell apoptosis, and inducing immune
   evasion [[114]60, [115]61]. CAFs communicate bidirectionally with tumor
   cells and other cells in the TME, mediating cell-derived factors,
   chemokines, growth factors, and exosomes. CAFs enhance tumor
   proliferation and induce immune escape in cancer cells, making them
   important anticancer targets [[116]62, [117]62, [118]63].

   We downloaded ESCC bulk RNA-seq data from the TCGA database in this
   study. Using the cell abundance scores calculated by MCPcounter and
   cell enrichment scores by xCell, we categorized samples into the high
   Fibroblasts group and the low Fibroblasts group based on the median
   Fibroblasts score. Differential analysis revealed 535 upregulated genes
   and 338 downregulated genes by MCPcounter; 1060 upregulated genes and
   262 downregulated genes by xCell (Fig. [119]1A–D). Growing evidence
   suggests that CAFs secreted chemokines influence tumor initiation and
   progression [[120]64]. We collected 49 human Chemokines [[121]65] and
   identified one upregulated gene (CXCL12) and three downregulated genes
   (CXCL1, CXCL8, CCL20) by intersecting Chemokines with MCPcounter and
   xCell results (Fig. [122]1E–F). Additionally, we have presented the
   expression of CXCL12, CXCL1, CXCL8, and CCL20 in MCPcounter and xCell
   analysis. The data indicated that CXCL12 expression in the Fibroblasts
   score High group was significantly higher than in the Fibroblasts score
   Low group, while CXCL1, CXCL8, and CCL20 expression levels were
   significantly lower in the Fibroblasts score High group. These results
   suggest a significant correlation between the increase in Fibroblasts
   score and the expression levels of these inflammation-related genes,
   implying that fibroblasts may play an important role in the regulation
   of these genes (Fig. [123]1G–H).

Fig. 1.

   [124]Fig. 1
   [125]Open in a new tab

   Analysis of ESCC-CAFs-related genes using bioinformatics tools. Note: A
   Heatmap of the top 100 differentially expressed genes related to
   Fibroblasts cell count in TCGA-ESCC analyzed by MCPcounter (Low: 41,
   High: 40); B Volcano plot of differentially expressed genes related to
   Fibroblasts cell count in TCGA-ESCC analyzed by MCPcounter (Red: high
   expression, Blue: low expression, Black: no difference); C Heatmap of
   the top 100 differentially expressed genes related to Fibroblasts in
   TCGA-ESCC analyzed by xCell (Low: 41, High: 40); D Volcano plot of
   differentially expressed genes related to Fibroblasts in TCGA-ESCC
   analyzed by xCell (Red: high expression, Blue: low expression, Black:
   no difference); E–F Venn diagrams showing the intersection of
   upregulated and downregulated genes obtained from Chemokines,
   MCPcounter, and xCell analyses; (G-H) Box plots of CXCL12, CXCL1,
   CXCL8, CCL20 expression in MCPcounter and xCell analyses (Fibroblasts
   score Low: n = 41, Fibroblasts score High: n = 40, * indicates p <
   0.05, ** indicates p < 0.01, *** indicates p < 0.001)

   Based on the above results, we conducted bioinformatics analysis on
   ESCC bulk RNA-seq data from the TCGA database and identified four
   chemokines (CXCL12, CXCL1, CXCL8, CCL20) closely associated with
   ESCC-CAFs.

Quality control, filtering, and PCA of scRNA-seq data from brain tissues of
ESCC patients

   To further investigate the role of CAFs in the development of ESCC, we
   obtained a scRNA-seq dataset of ESCC patient esophageal tissues through
   high-throughput sequencing. Using the Seurat package for data
   integration, we first examined the number of genes detected in each
   cell (nFeature_RNA), the total mRNA molecules detected (nCount_RNA),
   and the percentage of mitochondrial genes (percent. mt) in the
   scRNA-seq data. The results showed that most cells had nFeature_RNA
   < 7500, nCount_RNA < 10^5, and percent.mt less than 25%. In the ESCC2
   sample, percent.mt ranged from 20 to 75%, and nFeature_RNA and
   nCount_RNA were lower compared to other samples (Figure S3A). We
   filtered out low-quality cells using the criteria of nFeature_RNA > 50
   and percent.mt < 10, resulting in an expression matrix containing
   22,685 genes and 23,676 cells. Correlation analysis of sequencing depth
   showed a correlation coefficient of r = − 0.1 between nCount_RNA and
   percent. mt and r = 0.84 between nCount_RNA and nFeature_RNA in the
   filtered data, indicating good data quality for further analysis
   (Figure S3B). Subsequently, we performed gene expression variance
   analysis on the filtered cells, selecting the top 2000 highly variable
   genes based on variance for downstream analysis (Figure S3C).

   Before conducting PCA dimensionality reduction, standard preprocessing
   was performed. The RunPCA function performed PCA on the top 2000 highly
   variable genes, revealing no apparent batch effects between samples
   (Figure S3D). We displayed the major component genes of the first 6 PCs
   (Figure S4A) and generated a heatmap of the first 6 PCs using the
   DimHeatmap function (Figure S4B). Visualizing the significance of the
   PCs using the JackStrawPlot function showed that important PCs,
   typically with small p-values (above the dashed line), could
   effectively capture the information in the highly variable genes
   selected earlier (Figure S3E). Additionally, by combining the
   JackStrawPlot analysis with the ElbowPlot function, we selected the top
   9 PCs with a PC p value < 0.05 for subsequent t-SNE analysis (Figure
   S3F).

   Through quality control, filtering, and PCA of scRNA-seq data from
   brain tissues of ESCC patients, we obtained high-quality filtered cell
   data and selected the top 9 PCs for further analysis.

Cell clustering and analysis of cell communication unveiled the significant
involvement of the CXCL12-CXCR4 signaling axis in cell interactions

   We first analyzed the distribution of cells across different clusters
   in each sample (Figure S6A), which revealed distinct patterns between
   Normal and ESCC samples. Next, we performed t-SNE clustering analysis
   based on the top 9 PCs, resulting in the classification of all cells
   into 21 clusters (Figure S5A). We then identified the marker genes for
   each cluster and visualized the top 10 most highly expressed
   cluster-specific marker genes (Figure S5B). Finally, we compared the
   cluster distribution between Normal and ESCC samples, showing that
   Normal cells were mainly enriched in Clusters 0, 3, 4, 6, 8, 9, 10, 11,
   12, 13, 14, 15, 16, and 20, while ESCC cells were predominantly found
   in other clusters (Figure S5C). Using the Bioconductor/R package
   “SingleR” combined with the CellMarker database, we annotated the
   marker genes of different cell types in ESCC to the 21 cell clusters
   (Table S2), resulting in the annotation of 12 cell types (Mast cells,
   CD8^+ T cells, T cells, B cells, Fibroblasts, Macrophages, Monocyte,
   Stem cells, Endothelial cells, Monocyte, CAFs, Plasma cells). The
   quantities of different cell types between Normal and ESCC were
   significantly different, with Cluster#18 annotated as CAFs (Figure
   S5D).

   Comprehensive analysis based on samples, clusters, and cell types
   revealed the distribution of CXCL12, CXCL1, CXCL8, and CCL20
   expression. Results showed that CXCL12 was more prominently expressed
   in Fibroblasts (Cluster 4, 14), CAFs (Cluster 18), and Endothelial
   cells (Cluster 13); CXCL1 was mainly expressed in CAFs (Cluster 18) and
   Stem cells (Cluster 10); CXCL8 exhibited high expression in Macrophages
   (Cluster 8), Monocyte (Cluster 9), Stem cells (Cluster 10, 16), and
   Mast cells (Cluster 20); while CCL20 did not show significant
   distribution or expression in any cells (Figure S5E; Figure S6B-E).
   Both CXCL12 and CXCL1 showed high expression in CAFs.

   Subsequently, using the"CellChat"package, cell communication analysis
   was conducted, where gene expression data of cells were inputted, and
   interactions between ligands, receptors, and their auxiliary factors
   were simulated to explore the interactions between CXCL12 and CXCL1 in
   CAFs and immune cells and receptors in the TME. Initially, the CXCL
   signaling pathway-mediated cell interactions showed that CAFs exhibited
   the strongest interactions with CD8^+ T cells, T cells, and B cells
   (Figure S5F; Figure S6F-G). Further calculations on all
   ligand-receptor-mediated cell interactions and visualization of
   significant ligand-receptor pairs revealed a significant interaction
   between CAFs and CD8^+ T cells, CAFs and T cells, as well as CAFs and B
   cells mediated by the CXCL12/CXCR4 ligand-receptor pair (Figure S5G).
   Additionally, the receptor CXCR4 was prominently distributed and
   expressed in CD8^+ T cells (Cluster 1, 6), T cells (Cluster 2, 11, 12),
   and B cells (Cluster 3, 5, 7) (Figure S5H; Figure S5E; Figure S6H),
   consistent with the cell communication results. Furthermore, analysis
   of RNA-seq data from GTEx normal esophagus samples merged with ESCC
   RNA-seq data revealed that the expression levels of CXCL12 and CXCR4 in
   the tumor group were significantly higher than in the normal group
   (Figure S5I).

   In summary, we discovered a strong interaction between the CXCL12-CXCR4
   cell signaling axis in CAFs and CD8^+ T cells, T cells, and B cells in
   ESCC through single-cell clustering and cell communication analysis.

Overexpression of CXCL12 in human ESCC-CAFs

   Previous studies have shown that in pancreatic ductal adenocarcinoma,
   activated pancreatic stellate cells reduce the migration of CD8^+ T
   cells towards the tumor stroma through the CXCL12/CXCR4 axis, thereby
   preventing CD8^+ T cells from approaching cancer cells. The secretion
   of CXCL12 by pancreatic stellate cells can suppress effective
   anti-tumor immune responses [[126]66]; in ovarian cancer, the
   CXCL12-CXCR4 axis has also been confirmed to have an immunosuppressive
   effect [[127]67]. Furthermore, CXCL12 and its receptor CXCR4 are highly
   expressed in ESCC [[128]68]. Based on these findings, we will further
   investigate how CAFs regulate CD8^+ T cells via the CXCL12/CXCR4 axis
   to influence the development of ESCC. Firstly, we examined the
   expression of CXCL12 in human ESCC-CAFs, and the results from H&E
   staining and immunohistochemistry showed that CXCL12 is significantly
   expressed in the stroma of ESCC, while there is slight expression in
   the stroma of normal tissues (Fig. [129]2A). Immunofluorescence
   staining revealed that α-SMA, a marker protein for CAFs [[130]69], was
   co-expressed with CXCL12 in a significantly increased number of stromal
   cells in ESCC tissues. The markedly enhanced co-localization of α-SMA
   and CXCL12 in the ESCC stroma indicates high CXCL12 expression in CAFs
   within the TME (Fig. [131]2B).

Fig. 2.

   [132]Fig. 2
   [133]Open in a new tab

   Levels of CXCL12 expression in ESCC-CAFs. Note: A Hematoxylin and eosin
   staining and immunohistochemical testing were conducted to assess the
   pathological changes of ESCC clinical tissues and the expression levels
   of CXCL12 (50 μm and 25 μm magnification, Normal: n = 10, Tumor: n =
   10, * indicates p < 0.05 compared to the Normal group); B
   Immunofluorescence staining was used to detect the expression of α-SMA
   and CXCL12 in ESCC clinical tissues (50 μm, n = 10); C RT-qPCR was
   performed to detect the expression of α-SMA and FAPα in NFs and CAFs
   cells; D RT-qPCR was carried out to measure the levels of CXCL12
   expression in NFs and CAFs cells (* indicates p < 0.05 compared to the
   NFs group); E ELISA was utilized to determine the levels of CXCL12
   expression in NFs and CAFs cells (* indicates p < 0.05 compared to the
   NFs group). Each cell experiment was repeated three times

   To further explore the characteristics of CAFs and their interactions
   with tumor cells, we isolated CAFs and paired NFs from tumor tissue and
   adjacent non-tumor tissue, and conducted identification (Figure S1).
   RT-qPCR analysis of α-SMA and FAPα showed that the expression of α-SMA
   and FAPα is significantly increased in CAFs compared to NFs
   (Fig. [134]2C). RT-qPCR and ELISA results demonstrated that compared to
   the NFs group, the expression of CXCL12 is significantly increased in
   the CAFs group (Fig. [135]2D–E).

   These results indicate that CXCL12 is highly expressed in human
   ESCC-CAFs.

CAFs secrete CXCL12 to inhibit the tumor-killing activity of CD8^+ T cells

   Previous studies have shown that CD8^+ T cells are a critical component
   of anti-tumor immunity, and the success of immunotherapy relies on the
   activation of effective cytotoxic T-cell responses [[136]38].
   Therefore, we further investigated whether CAFs affect the infiltration
   of CD8^+ T cells by secreting CXCL12. Firstly, we constructed two
   CXCL12 silencing sequences and detected the silencing efficiency in
   CAFs cells isolated from ESCC tumors using RT-qPCR. The results showed
   that compared to the sh-NC group, the expression levels of CXCL12 were
   significantly decreased in the sh-CXCL12-1 and sh-CXCL12-2 groups
   (Fig. [137]3A). Since CXCL12 is a secretory protein, ELISA analysis of
   the supernatant of CAFs cells revealed that the CXCL12 concentration
   was also significantly reduced in the sh-CXCL12-1 and sh-CXCL12-2
   groups compared to the sh-NC group (Fig. [138]3B), with sh-CXCL12-1
   showing higher efficiency. Therefore, the sh-CXCL12-1 (sh-CXCL12)
   silencing sequence was used for subsequent experiments.

Fig. 3.

   [139]Fig. 3
   [140]Open in a new tab

   Impact of CXCL12 secreted by CAFs on the function of CD8^+ T cells.
   Note: A RT-qPCR was conducted to evaluate the silencing efficiency of
   CXCL12 in CAFs cells; B ELISA was utilized to measure the levels of
   CXCL12 in the supernatant of CAFs cells; C Schematic representation of
   co-culturing; D RT-qPCR was performed to detect the levels of CXCR4 in
   the co-culture system; E Western blot analysis was used to assess the
   expression changes of CXCL12 and CXCR4 in the co-culture system; F
   Immunofluorescence was performed to observe the chemoattractant
   capability of CAFs from various groups for CD8^+ T cells; G Flow
   cytometry was used to detect the changes in the expression of PD-1 and
   IFN-γ in CD8^+ T cells in the co-culture system, and quantitative
   results were obtained; H Flow cytometry was employed to measure the
   changes in the expression of GrzB and Perforin in CD8^+ T cells in the
   co-culture system; I Flow cytometry was used to quantify changes in
   PD-1 and IFN-γ expression in CD86^+ macrophages in the co-culture
   system; J Flow cytometry was also used to evaluate GrzB and Perforin
   levels in CD86^+ macrophages; K PD-1 and IFN-γ expression was assessed
   in CD56^+ NK cells; and L GrzB and Perforin expression was analyzed in
   CD56 + NK cells. *p < 0.05, ***p < 0.001; all cell experiments were
   replicated three times

   Next, we co-cultured CD8^+ T cells with CXCL12-silenced CAFs (GFP
   labeled) (Fig. [141]3C) and detected the expression of CXCL12 and its
   receptor CXCR4 in the co-culture system. The results showed that
   compared to the CAFs-sh-NC group, the expression of CXCL12 and CXCR4
   was significantly reduced in the CAFs-sh-CXCL12 group, while their
   expression levels were significantly elevated in the oe-CXCL12 group
   compared to the oe-NC group (Fig. [142]3D–E). As CXCR4 is the receptor
   for CXCL12, CXCL12 silencing leads to the downregulation of CXCR4 due
   to the absence of ligand binding, consistent with previous reports
   [[143]22, [144]70]. Immunofluorescence results demonstrated that the
   number of bound CD8^+ T cells in the CAFs-sh-CXCL12 group was
   significantly decreased compared to the CAFs-sh-NC group, and
   significantly increased in the oe-CXCL12 group compared to controls
   (Fig. [145]3F). We hypothesized that CAFs may recruit CD8^+ T cells
   through secreting CXCL12 to induce T cell exhaustion, thereby promoting
   immunosuppression. In tumor tissues, functional exhaustion of CD8^+ T
   cells is characterized by decreased IFN-γ and high expression of
   programmed cell death protein-1 (PD-1) [[146]71, [147]72]. Flow
   cytometry analysis of PD-1 and IFN-γ levels in CD8^+ T cells revealed
   that the expression of PD-1 was significantly reduced, while IFN-γ
   levels were significantly increased in the CAFs-sh-CXCL12 group
   compared to the CAFs-sh-NC group and IFN-γ levels were upregulated in
   CD8^+ T cells from the CAFs-sh-CXCL12 group, whereas the opposite trend
   was observed in the oe-CXCL12 group (Fig. [148]3G), further validating
   our hypothesis.

   To further evaluate the impact of this phenomenon on the tumor-killing
   function of CD8^+ T cells, we analyzed the expression of key cytotoxic
   markers, Granzyme B (GrzB), and Perforin in CD8^+ T cells co-cultured
   with CAFs [[149]73]. The results showed that compared to the CAFs-sh-NC
   group, the expression of Perforin and GrzB was significantly increased
   in the CAFs-sh-CXCL12 group. The number of CD8^+ T cells bound to CAFs
   also significantly increased in the oe-CXCL12 group versus the oe-NC
   group, indicating a substantial enhancement in the tumor-killing
   function of CD8^+ T cells (Fig. [150]3H). Similar results were observed
   in CD86^+ macrophages and CD56^+ NK cells (Fig. [151]3I–L).

   In conclusion, we found that in ESCC, CAFs secreting CXCL12 inhibit the
   tumor-killing effect of CD8^+ T cells, CD86^+ macrophages, and CD56^+
   NK cells, while silencing CXCL12 enhances their tumor-killing
   potential.

Validation of the promotion of ESCC growth and metastasis by CAFs through
CXCL12 secretion

   To further investigate the regulatory role of CAFs on ESCC, CAFs were
   co-cultured with TE-1 cells. ELISA results of TE-1 cells showed a
   significant increase in CXCL12 expression in the TE-1 + CAFs group
   compared to the TE-1 + NFs group. Additionally, there was a significant
   decrease in the expression of CXCL12 in the TE-1 + CAFs-sh-CXCL12 group
   compared to the TE-1 + CAFs-sh-NC group (Figure S7A). Subsequently,
   validation was conducted through CCK-8, EdU staining, flow cytometry,
   and Transwell assays. The results revealed that compared to the TE-1
   + NFs group, the TE-1 + CAFs group demonstrated significantly increased
   cell viability, proliferation, migration, and invasion, with reduced
   apoptosis. In contrast, the TE-1 + CAFs-sh-CXCL12 group exhibited
   significantly decreased cell viability, proliferation, migration, and
   invasion, with increased apoptosis compared to the TE-1 + CAFs-sh-NC
   group (Figure S7B-F). These in vitro experiments indicate that CAFs
   play a key role in promoting the growth and metastasis of ESCC cells
   through CXCL12 secretion.

   To further confirm the effect of CAF-secreted CXCL12 on tumor growth
   and metastasis in vivo, TE-1 cells were mixed with NFs or CAFs and
   injected into the axilla of CD34^+ humanized mice. The experimental
   results showed a significant increase in tumor volume and weight in the
   TE-1 + CAFs group compared to the TE-1 + NFs group. Furthermore, there
   was a significant reduction in tumor volume and weight in the TE-1
   + CAFs-sh-CXCL12 group compared to the TE-1 + CAFs-sh-NC group
   (Fig. [152]4A–C). Immunofluorescence (IF) analysis demonstrated a
   significant decrease in intratumoral CD8 T cells and activated CD8 T
   cells (GZMB +) in the TE-1 + CAFs group compared to the TE-1 + NFs
   group. Conversely, the TE-1 + CAFs-sh-CXCL12 group showed a significant
   increase in intratumoral CD8 T cells and activated CD8 T cells (GZMB +)
   compared to the TE-1 + CAFs-sh-NC group (Fig. [153]4D). Results from
   the lung metastasis model revealed a significant increase in lung
   metastatic foci in the TE-1 + CAFs group compared to the TE-1 + NFs
   group. Conversely, there was a significant reduction in lung metastatic
   foci in the TE-1 + CAFs-sh-CXCL12 group compared to the TE-1
   + CAFs-sh-NC group (Fig. [154]4E–F). These in vivo experiments
   demonstrate that CAF-secreted CXCL12 significantly promotes the growth
   and lung metastasis of ESCC cells.

Fig. 4.

   [155]Fig. 4
   [156]Open in a new tab

   Effects of CAF-secreted CXCL12 on the growth and metastasis of ESCC.
   Note: A Subcutaneous transplantation tumor changes in humanized mice
   with CD34^+ cells (n = 6); B Growth curve of subcutaneous
   transplantation tumors in humanized mice with CD34^+ cells (n = 5); C
   Variation in the weight of subcutaneous transplantation tumors in
   humanized mice with CD34^+ cells (n = 6); D Immunofluorescence analysis
   of CD8 T cell and activation quantity within the transplantation tumors
   of humanized mice with CD34^+ cells (red: CD8, green: GZMB, yellow
   arrow: GZMB inside CD8 T cells, scale bar: 25 μm, n = 6); E–F
   Metastatic foci in the lungs of humanized mice with CD34^+ cells and
   corresponding H&E staining images (yellow circle: metastatic foci,
   scale bar: 50 μm; n = 6). (* indicates p < 0.05 compared to the TE-1
   + NFs group, # indicates p < 0.05 compared to the TE-1 + CAFs-sh-NC
   group)

Construction of nano-spheres targeting CAFs for si-CXCL12 delivery

   Based on in vitro and in vivo experimental results, we observed that in
   ESCC, CAFs not only inhibit the tumor-killing function of CD8^+ T cells
   by secreting CXCL12 but also promote the proliferation and metastasis
   of ESCC cells. In light of this, we propose developing a nano-particle
   strategy targeting CAFs, where these nano-particles can deliver siRNA
   (si-CXCL12) to specifically target and knock down the expression of
   CXCL12 in CAFs. Through this approach, we aim to weaken the inhibitory
   effect of CAFs on CD8^+ T cell function, thereby enhancing their
   tumor-killing activity, while inhibiting the proliferation and
   metastasis of ESCC cells, providing a new therapeutic strategy for ESCC
   treatment.

   The preparation process involves constructing the core structure using
   materials such as SA and polyethyleneimine (PEI), activating carboxyl
   groups and coupling with PEI through EDC and NHS chemical reactions.
   Subsequently, the antibody Anti-FAPα is conjugated with
   DSPE-PEG2000-Mal through a Michael addition reaction to form the
   targeting molecule. This structure is consistent with the
   characteristics of polymeric lipid complexes, which enable drug loading
   within the core or on the surface [[157]74, [158]75]. Finally, these
   components are mixed in specific proportions, processed with ultrasound
   to form nano-particles, and complexed with CXCL12 siRNA to form
   LNCs@si-CXCL12 nano-particles (Fig. [159]5A). The DLS measurement shows
   the average diameter of LNCs@si-CXCL12 as 80.5 ± 2.8 nm (Fig. [160]5B),
   with a Zeta potential of 23.29 mV (Fig. [161]5C). TEM reveals that the
   nano-spheres have a spherical morphology with uniform size distribution
   (Fig. [162]5D). Including DSPE-PEG2000 enhances the colloidal stability
   of liposomes, providing them with long-circulating stealth properties.
   In PBS containing 10% FBS, the size of LNCs@si-CXCL12 remains unchanged
   at 37 °C for 72 h (Fig. [163]5E), indicating high stability under
   normal physiological conditions. Additionally, the release rate of
   LNCs@si-CXCL12 reaches 58% at 24 h (Fig. [164]5F). Gel electrophoresis
   further confirmed the systemic stability of the nanoparticles
   (Fig. [165]5G). In 3D multicellular tumor spheroid (MCTS) co-culture
   assays, LNCs@si-CXCL12 treatment significantly reduced the size of
   spheroids derived from TE-1 cells, whereas spheroids treated with blank
   LNCs continued to grow over five days. Moreover, LNCs-treated spheroids
   exhibited smooth, regular surfaces with dense cores and minimal
   peripheral apoptosis, while LNCs@si-CXCL12-treated spheroids had
   irregular surfaces and abundant apoptotic cells at the periphery,
   indicating potent cytotoxicity (Fig. [166]5H). Spectral data confirmed
   the successful synthesis of PSA using SA and PEI 1800 (Fig. [167]5I).
   Long-term stability assessments over three months showed no significant
   changes in particle size, polydispersity index, or zeta potential
   (Fig. [168]5J). The encapsulation efficiency remained above 95% at the
   end of the observation period (Fig. [169]5K). The LNCs@si-CXCL12
   nanoparticles also showed no significant changes in particle size or
   zeta potential during storage (Fig. [170]5L). Western blot analysis
   verified the successful covalent conjugation of Anti-FAPα to
   DSPE-PEG2000-Mal, as indicated by a clear shift in protein band size
   (Fig. [171]5M). Serum stability assays demonstrated that LNCs@si-CXCL12
   maintained structural integrity under physiological conditions
   (Fig. [172]5N).

Fig. 5.

   [173]Fig. 5
   [174]Open in a new tab

   Preparation and characterization of LNCs@si-CXCL12. Note: A Schematic
   illustration of LNCs@si-CXCL12; B DLS measurement of the size
   distribution of LNCs@si-CXCL12; (C) Zeta potentials of LNCs and
   LNCs@si-CXCL12; D TEM images of LNCs and LNCs@si-CXCL12, bar = 100nm; E
   Stability of LNCs@si-CXCL12 at 37 ℃ in PBS containing 10% FBS; (F)
   Release efficiency of LNCs@si-CXCL12 within 24 h; G Gel electrophoresis
   verified the systemic stability of conjugated nanoparticles; H 3D tumor
   spheroid co-culture assay demonstrated the anti-tumor and anti-fibrotic
   potential of LNCs@si-CXCL12; I Spectral analysis confirmed PSA
   synthesis using SA and PEI 1800; J Particle size and potential
   statistics, K encapsulation efficiency detection; L LNCs@si-CXCL12
   Particle size and potential statistics; M Western blot validated
   Anti-FAPα-DSPE-PEG2000-Mal conjugation; N Serum stability assay of
   nanoparticles. All experiments were repeated three times

   By successfully implementing the above experiments, we confirm that the
   nano-spheres LNCs@si-CXCL12 targeting CAFs exhibit good stability and
   uniformity under physiological conditions and show favorable targeting
   effects on tumor tissues.

CAFs exhibit excellent targeted uptake of LNCs@si-CXCL12 with low
cytotoxicity

   To further observe the targeted uptake of CAFs by LNCs@si-CXCL12, we
   co-cultured these nano-complexes with CAFs and evaluated the expression
   level of CXCL12 by RT-qPCR. The results showed that the expression of
   CXCL12 in CAFs decreased by approximately 65% after treatment, and this
   inhibitory effect persisted for at least 14 days (Fig. [175]6A),
   indicating that LNCs@si-CXCL12 effectively inhibits the transcription
   of CXCL12 in CAFs.

Fig. 6.

   [176]Fig. 6
   [177]Open in a new tab

   Targeted uptake and cytotoxicity of LNCs@si-CXCL12 by CAFs. A qRT-PCR
   detected CXCL12 transcript levels on days 1 and 14 following co-culture
   with LNCs@si-CXCL1; B and C Illustrating the uptake of LNCs@si-CXCL12
   by CAFs; D Uptake by 293 T cells was minimal (scale bar = 25 μm); E–G
   CCK-8 assay was performed to observe the impact of LNCs@si-CXCL12
   concentrations ranging from 0 to 100 mg/mL on the viability of CAFs,
   TE-1, and Het-1A cells. *** indicates p < 0.001 compared to the
   PBS/LNCs group. Cell experiments were repeated three times

   Given that LNCs@si-CXCL12 includes antibodies targeting the surface
   molecule FAPα of CAFs for targeting purposes, we used LNCs without
   adding FAPα antibodies as a control group. Both types of LNCs were
   labeled with Cy5.5 and co-cultured with CAFs and TE-1 cells. The
   results showed a significant increase in the uptake of FAPα-linked
   LNCs@si-CXCL12 in CAFs (Fig. [178]6B), while the uptake efficiency of
   LNCs and LNCs@si-CXCL12 was similar in TE-1 cells and 293 T cells
   (Fig. [179]6C–D), indicating a significant targeting effect of
   LNCs@si-CXCL12 on CAFs [[180]76, [181]77].

   Furthermore, we assessed the cytotoxicity of LNCs@si-CXCL12. Using the
   CCK-8 assay, we tested the impact of LNCs@si-CXCL12 at concentrations
   ranging from 0 to 100 mg/mL on CAFs, TE-1 cells, and human esophageal
   epithelial cells Het-1A. The results showed that after 6 h of
   treatment, the viability of all three cell types remained above 80%.
   However, CAFs exhibited slightly lower cell viability compared to TE-1
   and Het-1A cells, possibly due to the higher endocytic capacity of CAFs
   towards LNCs@si-CXCL12 (Fig. [182]6E–G).

   In summary, LNCs@si-CXCL12 demonstrates excellent targeting capability
   towards CAFs, effectively inhibiting CXCL12 expression without
   significantly impacting cell viability.

LNCs@si-CXCL12 inhibits tumor growth in spontaneous ESCC mice and enhances
immune response

   To further investigate whether LNCs@si-CXCL12 could attenuate the
   inhibitory effect of CAFs on CD8^+ T cell function, thereby enhancing
   their tumor-killing activity and inhibiting ESCC growth, we established
   a spontaneous ESCC mouse model by feeding with 4-NQO (Fig. [183]7A).
   LNCs@si-CXCL12 maintained stable knockdown of CXCL12 mRNA in tumor
   tissues up to day 14 post-treatment (Fig. [184]7B). Western blotting
   revealed increased expression of fibrosis-associated proteins in the
   ESCC group (Fig. [185]7C). At the 16th week of modeling, H&E staining
   results showed a relatively orderly structure of esophageal epithelial
   tissues without apparent abnormalities; by the 30th week, the
   esophageal epithelial structure became disordered, with cellular
   atypia, indicating the formation of intraepithelial neoplasia. However,
   after 2 weeks of treatment with LNCs@si-CXCL12, tissue slices in the
   32nd week showed decreased neoplastic cells, effectively inhibiting
   tumor progression (Fig. [186]7D).

Fig. 7.

   [187]Fig. 7
   [188]Open in a new tab

   Therapeutic effects of LNCs@si-CXCL12 on spontaneous ESCC mice. Note: A
   Schematic diagram of spontaneous ESCC mouse model induced by feeding
   with 4-NQO; B qPCR analysis of CXCL12 mRNA expression in CAFs; C
   Western blot analysis of fibrosis-related proteins in mouse tumor
   tissues; D H&E-stained images of esophageal tissues from mice at 16,
   30, and 32 weeks during modeling period (scale bar =  100 μm); E
   Statistical analysis of body weight changes in each group of mice; and
   Survival curve of mice in each group (n = 10); F Representative images
   of esophageal tissues from each group of mice and statistical analysis
   of tumor numbers; G–H Tumor volume and WBC count in each group; (I)
   Expression of Ki67 in tumor tissues of each group of mice (scale bar = 
   100 μm, magnified scale bar =  25 μm); J Immunofluorescence showing
   co-localization of αSMA and CXCL12 in tumor tissues (scale bar =  25
   μm); K Flow cytometry analysis of IFNγ and GzmB expression in CD8^+ T
   cells within tumor tissues of each group of mice. ** indicates p < 0.01
   compared to the PBS group, *** indicates p < 0.001

   Regarding monitoring mouse weight, the treatment with LNCs@si-CXCL12
   did not negatively impact mouse weight. Survival curve analysis
   revealed that compared to the LNCs group, LNCs@si-CXCL12 significantly
   prolonged the survival time of ESCC mice (p = 0.037) (Fig. [189]7E).
   Further analysis of esophageal tissues indicated a significant
   reduction in tumor numbers after treatment with LNCs@si-CXCL12
   (Fig. [190]7F), along with a significant decrease in the number of
   Ki67-positive cells within the tumors (Fig. [191]7G). Antitumor scoring
   criteria confirmed the therapeutic effect of LNCs@si-CXCL12 in vivo
   [[192]54, [193]78]. Furthermore, immunofluorescence analysis showed
   that LNCs@si-CXCL12 treatment significantly reduced the number of cells
   co-expressing α-SMA and CXCL12 within tumor tissues, indicating
   targeted silencing of CXCL12 in CAFs (F[194]ig. [195]7I). Tumor weight
   and WBC count were also markedly decreased (Fig. [196]7G–H).
   Corroborated by immunohistochemistry (Figure S8A). This further
   confirms the inhibition of tumor development in spontaneous ESCC mice
   by LNCs@si-CXCL12.

   Furthermore, by detecting the expression of the effector molecules IFNγ
   and GzmB produced by CD8^+ T cells in tumor tissues, we found that
   LNCs@si-CXCL12 treatment significantly increased the expression of
   these cytokines (Fig. [197]7J, K), indicating that the treatment
   stimulated the immune system's response against the tumor. Moreover,
   LNCs@si-CXCL12 treatment markedly decreased the expression of CXCL12
   and CXCR4 in tumor tissues, while VEGF levels remained unaffected
   (Figure S8B). Treatment with LNCs@si-CXCL12 did not cause morphological
   damage to vital organs such as the heart, liver, spleen, lungs, and
   kidneys in mice (Figure S9A), and the main biochemical indicators in
   mouse serum, including ALT, AST, BUN, and CR, remained within normal
   ranges after treatment (Figure S9B–E), indicating the safety of
   LNCs@si-CXCL12 treatment.

   To explore the underlying molecular mechanisms, transcriptome analysis
   was performed on tumor tissues before and after treatment. Volcano plot
   analysis (Fig. [198]8A) revealed that CXCL12 was significantly
   downregulated, while immune activation-related genes such as Ccl7 and
   Igfbp2 were significantly upregulated, suggesting enhanced immune
   responses in the TME. Moreover, genes involved in stromal remodeling
   and metastasis, including Pdgfrl and Fgfr2, were downregulated. A
   heatmap (Fig. [199]8B) further illustrated the distinct gene expression
   profiles between the control and treatment groups, with upregulation of
   immune-related genes (Ccl7, Igfbp2) and downregulation of tumor
   progression genes (CXCL12, Fgb, Pax5), as well as fibrosis-related
   genes (Pdgfrl, Serpina3k). Kyoto Encyclopedia of Genes and Genomes
   (KEGG) pathway analysis showed significant enrichment of pathways
   including chemokine signaling, complement and coagulation cascades, and
   cytokine–cytokine receptor interactions. Notably, differential
   expression of CCL7, CCL8, CCL12, CXCL12, and ACKR4 suggested enhanced
   immune infiltration. Changes in FGA, FGB, FGG, and SERPINA1A implicated
   a role in inhibiting tumor progression via coagulation regulation.
   Other pathways affected included PPAR signaling, insulin secretion,
   calcium signaling, and cytoskeletal regulation (e.g., TNNC1, TNNT3,
   MYH4) (Fig. [200]8C). Gene Ontology (GO) functional analysis revealed
   that LNCs@si-CXCL12 promoted immune infiltration via chemokine
   signaling and immunoglobulin-mediated immune response, while inhibiting
   adaptive ECM remodeling. Molecular function enrichment showed
   significant representation of chemokine activity, CCR receptor binding,
   and extracellular matrix structural constituents (Fig. [201]8D).

Fig. 8.

   [202]Fig. 8
   [203]Open in a new tab

   Transcriptomic analysis. Note: A Volcano plot displaying differentially
   expressed genes (DEGs) in tumor tissues after LNCs@si-CXCL12 treatment;
   B Heatmap showing hierarchical clustering of DEGs; C KEGG pathway
   enrichment analysis of DEGs; D GO functional enrichment analysis of
   DEGs

   Additionally, protein–protein interaction (PPI) network analysis and
   modular clustering using Cytoscape’s MCODE tool identified six
   significant functional modules (Figure S10A–F), mainly involved in TME
   regulation, immune response, metabolic modulation, and ECM remodeling,
   further validating the mechanistic role of LNCs@si-CXCL12 in tumor
   suppression (Fig. [204]9).

Fig. 9.

   [205]Fig. 9
   [206]Open in a new tab

   Schematic representation of the molecular mechanism by which CAFs
   upregulate CXCR4 expression via CXCL12 secretion to promote ESCC growth
   and metastasis

   In conclusion, LNCs@si-CXCL12 inhibits ESCC tumor growth and activates
   the tumor-killing function of CD8^+ T cells, demonstrating potential
   therapeutic value in suppressing ESCC development and enhancing tumor
   immune responses.

Discussion

   This study's findings reveal a positive correlation between CXCL12 and
   the quantity of CAFs in ESCC, while other chemokines such as CXCL1,
   CXCL8, and CCL20 demonstrate a negative correlation with the quantity
   and activity of CD8^+ T cells in ESCC. This discovery emphasizes
   chemokines’ diverse and complex roles in tumor development. Building
   upon this foundation, our research utilized scRNA-seqg technology to
   further elucidate the close interplay between the CXCL12-CXCR4 axis in
   CAFs and CD8^+ T cells, highlighting the pivotal role of this signaling
   axis in regulating the immune response in ESCC. Previous studies have
   shown that CXCL12 promotes the polarization and activation of M1-type
   macrophages, enhancing their cytotoxicity against pathogens and tumor
   cells, and regulates their pro-inflammatory responses and iNOS
   expression via CXCR4 binding (PMID: 34864921; 34847760). In addition,
   CXCL12 enhances NK cell activation and cytotoxicity, partly by
   modulating surface markers such as CD56 and NKp46 to promote tumor cell
   recognition (PMID: 34079127; 36781697). These results, together with
   our own, suggest that CXCL12 can reprogram the TME and exert complex
   immunomodulatory effects. Additionally, a targeted nanosphere,
   LNCs@si-CXCL12, was developed to efficiently deliver siRNA, selectively
   downregulating CXCL12 expression in CAFs, thereby diminishing the
   inhibitory effect of CAFs on CD8^+ T cell function and restoring or
   enhancing CD8^+ T cells'immune clearance capability against ESCC.
   Through both in vitro and in vivo experiments, this nanosphere has
   demonstrated significant inhibition of tumor cell proliferation and
   metastasis while enhancing the tumor-killing function of CD8^+ T cells,
   indicating its potential clinical utility.

   Notably, we employed single-cell RNA sequencing (scRNA-seq) to
   investigate the CXCL12–CXCR4 axis and revealed its tight interaction
   between CAFs and CD8⁺ T cells, underscoring the pivotal role of this
   signaling pathway in orchestrating the immune response within ESCC.
   This aspect of the study offers new insights into the complex crosstalk
   between tumor cells and the surrounding microenvironment. In parallel,
   we developed a CAF-targeting nanoparticle system, LNCs@si-CXCL12, which
   not only enables efficient siRNA delivery but also effectively
   downregulates CXCL12 expression in CAFs, thereby reversing the
   suppression of CD8⁺ T cell activity.

   Previous studies have demonstrated that reducing CXCL12 levels can
   enhance T cell function and improve the efficacy of immune checkpoint
   inhibitors such as anti–PD-1/PD-L1 antibodies [[207]79]. CXCL12, as a
   chemokine, primarily facilitates the recruitment of immunosuppressive
   cells including MDSCs and regulatory T cells (Tregs) into the TME. Its
   downregulation can diminish the infiltration of these suppressive
   cells, thereby facilitating effector T cell access to tumors and
   enhancing anti-tumor immunity [[208]80, [209]81].

   Comparison with existing literature reveals that CXCL12 exhibits
   similar regulatory patterns in various tumor types. For instance, high
   expression of CXCL12 in colorectal cancer correlates positively with
   patient survival rates and can stimulate cancer cell proliferation and
   metastasis [[210]82]. Similarly, overexpression of CXCL1 in oral
   squamous cell carcinoma is also associated with tumor proliferation,
   invasion, and metastasis [[211]83]. The CXCL12/CXCR4 axis exacerbates
   disease progression by promoting tumor cell migration and invasion,
   consistent with our observations of CXCL12's role in ESCC, where it
   enhances tumor invasiveness and metastatic potential by activating CAFs
   [[212]84, [213]85]. Various strategies have been proposed in research
   regarding the application of nanomaterials, especially in targeting
   CAFs for cancer therapy. For example, some studies have demonstrated
   the design of nanocarriers for specific targeting of CAFs, releasing
   chemotherapy drugs or RNA interference molecules to suppress the
   pro-tumor function of CAFs in the TME [[214]27, [215]28]. These studies
   validate the effectiveness of nanotechnology in precisely delivering
   therapeutic agents to complex TMEs. Our strategy using the
   LNCs@si-CXCL12 nanosphere delivery system targeting the downregulation
   of CXCL12 expression in CAFs is similar, but our approach emphasizes
   inhibiting tumor growth through immune microenvironment modulation.

   While this study provides novel insights into the role of CXCL12 in
   ESCC and demonstrates the application of nanomaterials in therapy,
   there are limitations. Firstly, although the results of in vitro and in
   vivo experiments are promising, the complexity of the TME may lead to
   uncertainty in clinical applications. Future research should explore
   the impact of other cell types and molecular mechanisms in the TME.
   Additionally, whether CXCL12 interferes with conventional chemotherapy
   or other immunotherapies remains unexplored and will be one of our next
   research directions. Secondly, this study primarily used mouse models
   rather than human clinical samples; therefore, the clinical relevance
   of the research findings needs validation through broader clinical
   trials. Conducting additional experiments in other ESCC cell lines
   would also improve the generalizability of our findings. Moreover,
   future investigations will focus on how LNCs@si-CXCL12 impacts other
   immune cell populations beyond CD8⁺ T cells. Lastly, given that tumor
   initiation and progression are multifactorial and multistep processes,
   future studies should further investigate the roles of other cells and
   molecular mechanisms in the growth and metastasis of ESCC. The use of
   multiplex spatial immunofluorescence, which retains spatial resolution
   at the cellular and subcellular level, will allow us to analyze the
   distribution and activation states of immune cells in the TME,
   providing deeper mechanistic insights into immune evasion strategies.
   Incorporating this technology will be a key focus in our future
   research. Regarding nanomedicine applications, previous studies have
   shown that nanoparticles can effectively target CAFs and modulate their
   pro-tumoral functions [[216]27, [217]28], consistent with our approach.
   However, unlike strategies that solely inhibit CAFs, we emphasize the
   importance of reprogramming the immune microenvironment. Our strategy
   focuses not only on suppressing the tumor-promoting properties of CAFs
   but also on reinvigorating immune effector cells to enhance tumor
   clearance. This immunomodulatory perspective provides new opportunities
   for the development of next-generation targeted immunotherapies.

   In conclusion, This study reveals a positive correlation between CXCL12
   and the number of cancer-associated fibroblasts (CAFs) in esophageal
   squamous cell carcinoma (ESCC), and that CXCL12 exerts an
   immunosuppressive effect by inhibiting CD8 + T cell function. This
   finding is consistent with existing studies, particularly regarding the
   role of the CXCL12/CXCR4 axis in immune evasion and metastasis in
   various cancers [[218]82, [219]84, [220]85]. We have explored the
   interaction between the CXCL12-CXCR4 axis and CAFs in regulating immune
   responses through single-cell RNA sequencing (scRNA-seq). Targeting
   CAFs with nanoparticles (LNCs@si-CXCL12) that effectively deliver siRNA
   and reduce CXCL12 expression has reversed the inhibition of CD8 + T
   cells, enhancing the tumor-killing effect. Previous studies have
   demonstrated the application of nanoparticles in targeting CAFs and
   modulating the immune microenvironment [[221]27, [222]28], and our
   approach further emphasizes strategies to enhance tumor clearance
   through immune activation, providing new insights for targeted
   immunotherapy. Overall, our study highlights the role of CXCL12 in ESCC
   and demonstrates the potential of targeting CXCL12 to improve immune
   responses, laying the foundation for future therapeutic strategies (Fig
   [223]9).

Supplementary Information

   [224]12951_2025_3476_MOESM1_ESM.jpg^ (1.7MB, jpg)

   Additional file 1. Figure S1. Identification of NF and CAFs. Note:
   Immunofluorescence was conducted to observe the expression of the
   fibroblast marker gene fibrillinectin and the endothelial cell marker
   gene Cytokeratin in NF/CAFs; Immunofluorescence was performed to
   observe the expression of the CAFs markers α-SMA and FAPα in NF/CAFs
   [225]12951_2025_3476_MOESM2_ESM.jpg^ (586.1KB, jpg)

   Additional file 2. Figure S2. Purity Detection of CD8+ T Cells. Note:
   Flow cytometry detection of the purity of the extracted Jurkat cells;
   Flow cytometry was used to assess the purity of CD8+ T cells extracted
   from mice
   [226]12951_2025_3476_MOESM3_ESM.jpg^ (561.4KB, jpg)

   Additional file 3. Figure S3. scRNA Sample Analysis and Filtering.
   Note: Violin plots displaying the gene count, mRNA molecule count, and
   mitochondrial gene proportion of all cells in each sample; Scatter
   plots showing the correlation between nCount_RNA and percent.mt,
   nCount_RNA and nFeature_RNA; Identification of highly variable genes
   among 23676 genes through variance analysis; Scatter plot of cell
   clustering in samples analyzed by PCA; p-values for the top 40 PCs
   determined by PCA; Elbow plot analysis determining the use of the top 9
   PCs for t-SNE clustering analysis
   [227]12951_2025_3476_MOESM4_ESM.jpg^ (3.6MB, jpg)

   Additional file 4. Figure S4. PCA of scRNA-seq Data. Note: PCA of
   feature genes in the first 6 PCs and heatmap showing the expression
   levels of feature genes
   [228]12951_2025_3476_MOESM5_ESM.jpg^ (3MB, jpg)

   Additional file 5. Figure S5. Cell Clustering and Cell Communication
   Analysis. Note: t-SNE clustering analysis categorizing cells into 21
   cell clusters; Heatmap of the top 10 marker gene expressions for each
   cell cluster; t-SNE clustering analysis showing the clustering of
   Normal and Tumor samples; Annotation of 21 cell clusters into 12 cell
   types using the"SingleR"package in combination with the Cell Marker
   database; Distribution of CXCL12, CXCL1, CXCL8, and CXCR4 in cell
   clusters; Cell interaction network diagram mediated by CXCL signaling
   pathways; Bubble plot of all ligand receptor-mediated cell
   interactions, with a red box indicating a significant cell interaction
   relationship between CXCL12/CXCR4-mediated CAFs and CD8+ T cells, T
   cells, and B cells; Distribution of CXCR4 in the clustering plot;
   Expression patterns of CXCL12 and CXCR4 in normal and tumor groups; ***
   indicates P < 0.001 compared to the normal group
   [229]12951_2025_3476_MOESM6_ESM.jpg^ (2.4MB, jpg)

   Additional file6Figure S6. Cell Clustering and Cell Communication
   Analysis. Note: t-SNE cluster analysis of 6 samples; Distribution of
   CXCL12, CXCL1, CXCL8 in the cluster map; Distribution of CXCL12, CXCL1,
   CXCL8 in 21 cell clusters; Cell interaction chord diagram and
   hierarchical diagrammediated by CXCL signaling pathway; Distribution of
   CXCR4 in 21 cell clusters
   [230]12951_2025_3476_MOESM7_ESM.jpg^ (6.2MB, jpg)

   Additional file 7. Figure S7. Effect of CAFs on the Growth and
   Metastasis of ESCC by Secreting CXCL12. Note: ELISA to detect the
   expression levels of CXCL12 in TE-1 cells in each treatment group;CCK-8
   assay to evaluate the viability of TE-1 cells in each treatment group;
   EDU assay to measure the proliferation of TE-1 cells in each treatment
   group; Flow cytometry to detect the apoptosis of TE-1 cells in each
   treatment group; Transwell assay to assess the migration and invasion
   of TE-1 cells in each treatment group. * indicates P < 0.05 compared to
   TE-1+NFs group, # indicates P < 0.05 compared to TE-1+CAFs-sh-NC group.
   All cell experiments were performed in triplicate
   [231]12951_2025_3476_MOESM8_ESM.jpg^ (1.4MB, jpg)

   Additional file 8. Figure S8. Targeting Effect of LNCs@si-CXCL12 on
   CAFs. Note: Immunohistochemistry to observe the alteration in the
   expression of the CAF activation marker αSMA in tumor tissues of mice;
   Western blot analysis to detect the expression of CXCL12,,VEGF and
   CXCR4 in tumor tissues of different groups. ** indicates P < 0.01
   compared to the PBS group, *** indicates P < 0.001.
   [232]12951_2025_3476_MOESM9_ESM.jpg^ (2.7MB, jpg)

   Additional file 9. Figure S9. Biological Safety Evaluation of
   PAMAM@Cur. Note: H&E stained tissue sections of mouse heart, liver,
   spleen, lung, and kidney for morphological changes post PAMAM@Cur
   treatment; Evaluation of biochemical parameters ALT, AST, BUN, and CR
   in mouse serum post-treatment to assess the biocompatibility and safety
   of PAMAM@Cur. n=6.
   [233]12951_2025_3476_MOESM10_ESM.jpg^ (1MB, jpg)

   Additional file 10. Figure S10. PPI Network Modules. Note: Modular
   clustering of the PPI network using the Cytoscape MCODE plugin,
   identifying six prominent functional clusters related to TME
   regulation, immune responses, metabolic pathways, and extracellular
   matrix remodeling
   [234]Additional file 11^ (13.4KB, docx)
   [235]Additional file 12. ^ (12KB, docx)
   [236]Additional file 13. ^ (12.6KB, docx)

Acknowledgements