Abstract

   The Triple-Negative Breast Cancer (TNBC) subtype constitutes 15-20% of
   breast cancer cases and is associated with the poorest clinical
   outcomes. Distant metastasis, particularly to the lungs, is a major
   contributor to the high mortality rates in breast cancer patients.
   Despite this, there has been a lack of comprehensive insights into the
   heterogeneity of metastatic tumors and their surrounding ecosystem in
   the lungs. In this study, we utilized spatial RNA sequencing technology
   to investigate the heterogeneity of lung metastatic tumors and their
   microenvironment in two spontaneous lung metastatic mouse models. Our
   findings revealed an increase in metabolic-related genes within the
   cancer cells, with the hub gene Dlat (Dihydrolipoamide
   S-Acetyltransferase) showing a significant association with the
   development of lung metastatic tumors. Upregulation of Dlat led to the
   reprogramming of fatty acid utilization, markedly enhancing the
   bioenergetic capacity of cancer cells. This finding was corroborated by
   the increased dependence on fatty acid utilization in lung metastatic
   cancer cells, and inhibition of Dlat in breast cancer cells exhibited a
   reduced oxygen consumption rate. Consequently, inhibition of Dlat
   resulted in decreased survival capacity of breast cancer by reducing
   cancer stem cell properties and cell adhesion in the lung in vivo. The
   three cell components within the lung metastatic niche, including
   CD163^+ macrophages, neutrophils, and endothelial cells, expressed
   elevated levels of ApoE, leading to the secretion of various
   protumorigenic molecules that promote cancer cell growth in the lung.
   These molecules include galectin-1, S100A10, S100A4, and S100A6.
   Collectively, our findings highlight the lipid metabolism reprogramming
   of cancer and components of the tumor microenvironment that support
   lung metastasis of TNBC breast cancer.

   Subject terms: Cancer metabolism, Transcriptomics

Introduction

   Breast cancer (BC) is currently one of malignant and is estimated to
   account for 30% of all new cancer diagnosed in women. It is also the
   major cause of most cancer-related deaths, and metastasis of breast
   cancer to vital organs is considered the principal factor
   [[46]1–[47]3]. Therefore, BC remains a vital public health concern,
   particularly triple-negative breast cancer (TNBC), which is the absence
   of well-defined and universal targets for therapy, resulting in the
   poorest prognosis among breast cancer patients [[48]4]. The lung is the
   second most common site of breast cancer metastasis, with a 5-year
   overall survival of approximately 16.8% [[49]3–[50]5]. The poor
   clinical outcome of TNBC patients with lung metastasis is attributed to
   the limited options of therapeutic strategy accompanying inoperable
   lesions [[51]6]. Improving our understanding of the molecular mechanism
   of lung metastasis in BC might open novel diagnostic and new
   therapeutic approaches for amending the prognosis and treatment of
   advanced breast cancer.

   Heterogeneity typically exists between similar tumor types, resulting
   in subtypes (intertumor heterogeneity) or within tumors of the same
   type (intratumor heterogeneity). Intratumor heterogeneity endows cancer
   cells with different oncogenic properties and resistance to treatment;
   thus, they exhibit different abilities to cope with various challenges,
   such as new microenvironment adaption during metastasis [[52]7–[53]9].
   The tumor microenvironment (TME) is a significant regulator of cancer
   progression. The components of TME, such as cancer fibroblast
   populations, tumor vasculature, immune system infiltration, and
   extracellular matrix remodeling, may cause different tumor
   heterogeneity [[54]10, [55]11]. In addition to TME of the primary site,
   the metastatic niche is also thought to change before/during cancer
   spreading, and it also has a significant influence on the determination
   of cancer cell fate when disseminated cancer cells arrive at distant
   sites [[56]12–[57]14]. Conceivably, the phenotypic difference between
   cancer cells and distinct TME eventually determined the development of
   tumor heterogeneity in metastatic niches [[58]15, [59]16].
   Comprehensive profiling of the metastatic lung TME and its influence on
   breast cancer development has seen few, therefore, systematic and
   deepened research on the molecular heterogeneity of metastatic breast
   cancer and lung niche would upgrade the ability to explore more
   effective metastasis-targeting agents and improve the prognosis of BC
   patients with lung metastasis.

   Spatial transcriptomics (ST) and proteomics technologies (and
   proteomics), based on either RNA sequencing or imaging, provide an
   unprecedented opportunity to generate innovative insights into gene
   regulation in the spatial tissue context. ST offers spatial information
   and is a powerful tool to characterize genetic and functional
   heterogeneities [[60]17, [61]18]. Here, we use ST and multiple-staining
   imaging systems to explore the heterogeneity of metastatic breast
   cancer and lung niche using a TNBC animal model. We leverage two
   spontaneous lung metastatic models, TNBC 4T1 and MMTV-PyVT, to
   recapitulate spontaneous lung metastasis to investigate the tumor
   complexity in vivo, and the results reveal that fatty acid utility
   triggered metabolism-related genes required for lung spreading in
   breast cancer. Gene enrichment analysis and ligand-receptor pairs
   coupled to functional approaches implicate macrophage-derived soluble
   factors contributed to the remodeling of TMEs and tumor aggressiveness.
   Our work describes a functional interaction between TMEs and the
   malignant states that manipulate cancer progression during distant
   metastasis. We propose that increased fatty acid utility acts as a hub
   transition of cancer cells and immune responses of TME control BC lung
   metastatic fate.

Materials and methods

Cell culture

   Breast cancer cell lines Hs578T and 4T1 and human monocytic leukemia
   cell line THP-1 were obtained from American Type Culture Collection.
   Breast cancer cell lines were grown in DMEM supplemented with 10% Fetal
   Bovine Serum (FBS), while THP-1 were cultured in RPMI 1640
   (ThermoFisher, USA) medium with 10% FBS. All cell lines authenticated
   by short tandem repeat analysis. Each cell line was confirmed negative
   for mycoplasma contamination using mycoplasma test kits (Mycoalert
   Mycoplasma Detection Kit; Capsugel; Lonza Group, Ltd.) every 3 months.
   THP-1 monocytes are differentiated into macrophages by treating the
   cells with 150 nM phorbol 12-myristate 13-acetate (PMA, Sigma, P8139)
   for 24 h and then incubated in RPMI 1640 medium for an additional 24 h
   to allow for full differentiation into macrophages. THP-1-derived
   macrophages and Hs578T cells were co-cultured in a transwell system
   using RPMI 1640 medium for 16 h. In the blocking experiments, Hs578T
   cells were treated with or without etomoxir (4 μM) for 24 h. After
   washing, Hs578T cells were co-cultured with THP-1-derived macrophages,
   as described above.

Animal model

   All mice procedures related to animal use were conducted and approved
   in accordance with the Institutional Animal Care and Use Committee at
   Kaohsiung Medical University, and all animals were treated in
   accordance with the National Laboratory for Experimental Animals
   guidelines. BALB/c mice (female, 8-week old) received wild-type, or
   Dlat-knockdown/knockout 4T1 cells implantation (500,000 cells per
   mammary fat pad), allowing cells to metastasize to the lung for 21 days
   spontaneously. Female and male MMTV-polyoma middle T antigen (PyMT)
   transgenic mice on an FVB background (double transgenic; FVB/N-Tg
   [MMTV-PyVT] 634 Mul/J^−1) were obtained from Jackson Laboratories (Bar
   Harbor, ME, USA). A breeding colony was established by crossing FVB/NJ
   females with hemizygous MMTV-PyVT 634 Mul/J^−1males. These mice began
   developing palpable mammary tumors at around 4–5 weeks of age, with
   lung metastases appearing at approximately 15 weeks. Lungs of mice were
   harvested after the indicated times, fixed and embedded in paraffin,
   and sectioned for ST or CODEX ((CO-Detection by indEXing) analysis. For
   cell sorting, DiR-staining 4T1-L or lung metastatic breast cancer
   (MMTV-L, isolated from lung of MMTV-PyVT mice) cells were transplanted
   into BALB/c mice from mammary fat pads or tail vein injection. The
   animals were then sacrificed on 24 h (cell adhesion), 7 days (cell
   proliferation), and 21 days (tumor nodules). The ability of cell
   adhesion and proliferation in the lungs of mice was assessed by
   counting DiR-staining 4T1 or MMTV-L cells in a frozen section of the
   lung by a laser scanning confocal microscope (LSM700, Carl Zeiss,
   Oberkochen, Germany). The 4T1 breast cancer, macrophage cells, and
   CD11b^+F4/80^+ macrophage in the lungs were isolated by mincing,
   collagenase type I digestion, and sorting using cell sorter BD FACS
   Melody (BD Biosciences).

Slide preparation and ST sequencing

   Visium Spatial Gene Expression Slides (Visium Spatial Gene Expression
   Slide Kit) were obtained from 10x Genomics (Pleasanton, CA, USA), and
   they contained capture areas (6.5 mm^2) with approximately 5000
   barcodes and gene expression spots. The RNA integrity of formalin fixed
   & paraffin embedded (FFPE) sections were evaluated using the Agilent
   4150 TapeStation system. RNA quality. The lung FFPE blocks were cut
   into 10 mm-thickness tissue blocks and attached to the capture area of
   slides and then fixed and sequentially stained by Mayer’s hematoxylin
   (Millipore Sigma, MHS16) for 3 min, bluing buffer (cat no, 6697, Thermo
   Scientific) for 1 min, and eosin (cat no, HT110116, Millipore Sigma)
   for 1 min. The FFPE slide images were documented at 20× magnification
   using a TissueFAX (Tissue Gnostics, Vienna, Austria). The slides were
   permeabilization, and the mRNA released from lung sections during
   permeabilization, and the cDNA library was established according to the
   Visium spatial gene expression user guide. The cDNA library was
   sequenced on a NextSeq 2000 Systems (Illumina) with a sequencing depth
   of approximately 250‒270 M reads for each sample. (San Diego, CA, USA).

Raw sequencing data processing

   The Spatial Transcriptomics and bright-field H&E staining images were
   analyzed and merged by Space Ranger (version 1.1.0, 10x Genomics
   (Pleasanton, CA, USA), align reads, generate feature-spot matrices, and
   perform clustering and gene expression analysis. These pipelines
   combined Visium-specific algorithms with the widely used RNA-seq
   aligner STAR.

Weighted gene co-expression network analysis (WGCNA)

   The accessible R package of scWGCNA (v0.0.0.9) for ST was used for
   co-expression analysis ([62]https://github.com/smorabit/hdWGCNA). The
   Pearson Correlation Coefficient was used to assess similarities between
   paired mRNAs based on gene expression matrices. These similarities were
   then converted into adjacency matrices, where a soft-thresholding power
   was applied to emphasize strong gene connections while minimizing
   weaker correlations. The adjacency matrix was transformed into a
   topological overlap measure (TOM) to better capture the strength of
   connections between genes. Hierarchical clustering was performed using
   the TOM as input, allowing for the identification of network modules
   through the DynamicTreeCut method. Modules with high similarity were
   subsequently merged based on a specified threshold value. The module
   eigengene expression, adjacency matrix heat-map, and other related
   parameters were calculated and visualized.

Bioinformatics analysis of differential expression genes (DEG) analysis, Gene
Set Enrichment Analysis (GSEA), and trajectory analysis

   Signature genes needed to be expressed, and the changes in gene
   expression were > 2-fold (log2 FC > 1, p-value < 0.05). Pathway
   enrichment analysis in specific spots was performed using the KEGG
   database via the DAVID website ([63]https://david.ncifcrf.gov). This
   revealed key pathways and enriched biological processes related to gene
   expression in those regions, offering insights into underlying
   mechanisms (p < 0.05). Trajectory analysis was executed using Monocle 2
   (v2.22) to determine cell differentiation pathways and progression
   states. CytoTRACE was utilized to assess the transcriptional diversity
   of malignant cells based on their differentiation or stemness status
   ([64]https://cytotrace.stanford.edu). Cells were assigned a CytoTRACE
   score reflecting their differentiation potential, where higher scores
   indicate greater stemness and less differentiation. Pathway enrichment
   analyses were completed using the ssGSEA with default parameters gene
   sets were acquired from the MSigDB database
   ([65]https://www.gsea-msigdb.org/gsea/msigdb/).

Multiplex-CODEX staining and imaging of the lungs of mice

   Barcode-conjugated antibody staining of tissue sections mounted on
   poly-lysine coated coverslips was performed using a commercially
   available CODEX Staining Kit for FFPE tissue using a commercial Akoya
   CODEX instrument (Akoya Biosciences, Marlborough, MA) with a Keyence
   BZ-X800 microscope plus Nikon PlanApo NA 0.75 objective. Images were
   acquired at 20X magnification and filter cubes for TRITC (550), CY5
   (647), and CY7 (750) complementary fluorescent reporters and DAPI
   (Akoya Bio-sciences). Typical images are 3 × 3 mm, including
   acquisition of 5 Z-stack images. The images and the quantitation of
   specific cell types were collected using the CODEX Multiplex Analysis
   Viewer (MAV) software. The antibodies used in this study are listed in
   Supplementary Table [66]1.

Target genes knockdown or knockout by shRNA or siRNA transfection, and
CRISPR/Cas9

   4T1 were transfected with control shRNA or Dlat shRNA, and the stable
   clones were established by puromycin selection. The knockdown efficacy
   of Dlat was confirmed by quantitative real-time reverse transcription
   polymerase chain reaction assay (qRT-PCR) and Western blot, as
   described below. 4T1-L and MMTV-L cells (MMTV cells isolated from lung
   of MMTV-PyVT mice) cells were co-transfected with a Dlat-specific gRNA
   vector (Cat# [67]KN504604, OriGene Technologies) using the Turbofectin
   transfection reagent. Two distinct gRNA vectors were employed,
   containing Dlat-targeting sequences 5′-GGGGGATCCCACTACCGCAA-3′ (gRNA1)
   and 5′- CGCTTGCTGCGACAACTCCT-3′ (gRNA2). Stable clones of both 4T1-L
   and MMTV-L cells were analyzed by genomic PCR with primers flanking the
   predicted target sites in the Dlat gene. The primers used were
   CTTTTACGACACTGCGCGAC and TTCAGCAAAGCCACTCCCTT.

   ApoE knockdown was performed using Accell siRNA technology (Dharmacon).
   THP-1 cells were grown in Accell siRNA Delivery Medium and incubated
   with control or Accell™ siRNA (1 μM, Dharmacon) specifically targeting
   ApoE. The siRNA was added directly to the culture medium without the
   need for additional transfection reagents, following the manufacturer’s
   instructions. Cells were incubated for 48 h to allow for efficient gene
   silencing. Knockdown efficiency was confirmed via quantitative
   real-time PCR.

qRT-PCR

   The total RNA of the 4T1 or THP-1 cells was extracted and cDNA was
   converted form RNA using RT Reagent Kit (RR037A, TaKaRa), and 50 ng
   cDNA was used with Fast SYBR Green Master Mix (4385612, Applied
   Biosystems), followed by reactions using a QuantStudio 3 machine
   (Applied Biosystems). Gapdh was used as the internal control gene to
   give the tested genes a relative fold change using the 2^−ΔΔCt method.
   The qRT-PCR primers are listed in SupplementaryTable [68]2.

Western blot

   Western blot analysis is used to detect the expression levels of
   specific proteins. First, cells were collected, and total protein is
   extracted using a RIPA lysis buffer. Protein concentration is
   determined using the bicinchoninic acid assay. Total protein was
   separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis,
   followed by transfer onto a polyvinylidene difluoride membrane. The
   membrane is blocked with 5% milk and then incubated with a primary
   antibody overnight, followed by incubation with a horseradish
   peroxidase-conjugated secondary antibody for 1 h. The membrane is then
   washed, and protein bands are visualized using enhanced
   chemiluminescence detection reagents and an imaging system. The
   antibodies are listed in Supplementary Table [69]1

Cell functional assays

   The cell proliferation of 4T1 or Dlat knockdown was assessed by
   Premixed WST-1 Cell Proliferation Reagent (Clontech, Mountain View, CA)
   after incubation 24, 48, and 72 h. The migration of cells was assessed
   by wound healing and transwell (pore size: 0.8 μm) migration after 24 h
   incubation. Control plasmid transfected 4T1-L, Dlat knockdown/knockout
   4T1-L cells, control plasmid transfected MMTV-L or Dlat-knockout MMTV-L
   were seeded in a low-attached dish for 14 days, and the feature of
   cancer stem cells (CSCs) in cancer spheroids was determined by an
   ALDEFLUOR™ Assay Kit (StemCell technologies, Durham, NC, USA) with a BD
   Accuri™ flowcytometry (BD Bioscience, San Jose, CA, USA). The cells
   were stained with DEAB cocktail (DEAB reagent; 5 μl ALDF reagent;
   2.5 μl ALDF buffer; 500 μl/test) as the negative control.

Cell metabolic assay

   Control cDNA transfected 4T1-L or Dlat knockdown 4T1-L cells (6 × 10^3)
   or breast cancer cells isolated from lungs/primary sites (4T1-L and
   MMTV-L) or primary sites (4T1-P and MMTV-P) of mice were seeded in XF
   96-well and cultured in media in XF HS mini plate. The XF Cell Mito
   Stress Test kit (Cat no. 103010-100) and Mito fuel flex test (Cat no.
   1032870-100) were used following the manufacturer’s instructions to
   assay fuel utility and mitochondrial metabolism of the Dlat-knockdown
   4T1-L or MMTV-L cells.

Immunohistochemistry (IHC)

   Lungs of the mice were analyzed using IHC with anti-Dlat (Proteintech,
   13426-1-AP, 1:50), estrogen receptor, progesterone receptor, and HER2
   antibodies. Data were captured using Tissue FAXS (TissueGnostics,
   Vienna, Austria) based on the IHC staining slides. The antibodies used
   for IHC were listed in Supplementary Table [70]1.

Flow cytometry analysis of surface marker expression

   THP-1-derived macrophages were incubation with monoclonal antibodies or
   isotype-matched negative controls for 60 min at 4 °C in the dark. For
   intracellular staining, cells were fixed using Cytofix and
   permeabilized with Perm/Wash buffer, according to the manufacturer’s
   guidelines. Stained cells were then analyzed on a BD Accuri™
   flowcytometry (BD Bioscience, San Jose, CA, USA), with a minimum of
   10,000 events recorded per sample. The fluorescence-conjugated
   anti-human antibodies were listed in Supplementary Table [71]1.

Statistical analysis

   Data analysis was performed using Prism 9.2.0 (GraphPad Software).
   Unpaired t-tests were employed to compare two groups, while one-way
   ANOVA was utilized for comparisons involving more than two groups. The
   results presented in this study were indicative of three independent
   experiments, demonstrating consistent outcomes. All data are expressed
   as mean ± standard deviation (SD), and statistical significance was
   defined as a p-value below 0.05.

Results

Spatially mapping heterogeneity of metastatic lung nodules

   To investigate the spatial organization of metastatic lung nodules and
   their microenvironment, we performed spatially resolved transcriptomics
   on the lungs of mice with two spontaneously metastatic lung nodules in
   4T1 (TNBC) and MMTV-PyVT luminal androgen receptor-positive TNBC mouse
   models [[72]19, [73]20] (Fig. [74]1A and Supplementary Fig. [75]1A). A
   pathologist examined and annotated the sections based on the morphology
   of the associated H&E staining and visualized by t-distributed neighbor
   embedding (t-SNE) plots. Regions were labeled as Cancer-A1 and
   Diffusive cancer-A (in A section), Cancer-B1 to Cancer-B4 (B section),
   Diffusive cancer-B (B section), bronchial cells, and lung normal region
   in 4T1 mouse model (Fig. [76]1B and Supplementary Fig. [77]1B, [78]C);
   Meanwhile, in the MMTV-PyVT mouse model, regions in the lung and
   mammary gland were labeled as Cancer-C1 to C5, Primary-C1 and
   Primary-C2, respectively (Fig. [79]1C). The tumor microenvironments
   (TMEs) were classified as Cancer-A1-TME and Cancer-B1/2/3-TME,
   Cancer-B4-TME, and Diffusive cancer-TME in the 4T1 mouse model. TME in
   the MMTV-PyVT mouse model was assigned as Cancer-C-TME.

Fig. 1. Spatial transcriptomics (ST) of metastatic lung nodules in the 4T1
mouse model.

   [80]Fig. 1
   [81]Open in a new tab

   A Experimental workflow. Lung sections from two mouse models were
   utilized to generate spatial data. H&E stain, t-SNE representation, and
   spatial distribution were identified by clustering the integrated ST
   dataset across sections from the lungs of the 4T1 model (n = 6 for
   spontaneous metastasis, n = 2 for ST) (B) and the MMTV-PyVT (C) models
   (n = 2 for ST). Malignant status of different tumor nodules in the lung
   of 4T1 model (D) and MMTV-PyVT (E) model. CytoTRACE analysis of varying
   tumor nodules in the lung of the 4T1 model (F) and the MMTV-PyVT model
   (G). H The survival rate of breast cancer expressing the top 100 genes
   of aggressive cancer nodules (Cancer-B1). OS, overall survival; PSF,
   progression-free survival; DSS, disease-specific survival; DFI,
   disease-free interval.

   To further determine the degree of tumor aggressiveness, we calculated
   the oncogenic score based on expressed gene sets (hypoxic, mesenchymal,
   proliferation, cancer stem cells (CSCs), and epithelial phenotype). The
   results showed that Cancer B1 and Cancer B3 were the most aggressive,
   exhibiting the highest hypoxic, mesenchymal, and CSC scores, whereas
   diffusive cancer exhibited the lowest score in the 4T1 mouse model
   (Fig. [82]1D, Supplementary Fig. [83]1D). Similarly, Cancer-C3 in the
   MMTV-pyVT model was the most aggressive phenotype (Fig. [84]1E). We
   applied CytoTRACE to define the degree of tumor aggressiveness based on
   the number of expressed genes [[85]21]. Consistent with the biological
   score, Cancer B1 and Cancer C3 had the lowest degree of differentiation
   in TNBC mouse models, while diffusive cancer cells exhibited high
   differentiation (Fig. [86]1F, G). Gene Set Variation Analysis (GSVA)
   also revealed that the higher score of the top 100 genes of Cancer B1
   was associated with poor disease-specific survival (DSS) in breast
   cancer patients using the TCGA cohort (Fig. [87]1H).

The alternative of cell metabolism in lung metastatic tumor nodules

   To elucidate the spatiotemporally specific transcriptional network and
   identify key factors in lung metastasis, we conducted Weighted Gene
   Co-Expression Network Analysis (WGCNA). WGCNA identified two gene
   modules (Module 1 (M1) and Module 2 (M2)) associated with the degree of
   tumor aggressiveness in the 4T1 mouse model, as shown in Supplementary
   Fig. [88]2A–[89]C. KEGG pathway analysis revealed that M1 was linked to
   “Pathways in cancer,” “Mitophagy,” “Apoptosis,” and the “Hippo
   signaling pathway.” It exhibited higher activity in tumor nodules
   (Supplementary Fig. [90]2D). The expressions of M2 genes, encoding
   “Antigen processing and presentation”, “Cell cycle-related pathway,”
   and several metabolic functions, including “Carbon metabolism”, “TCA
   cycle”, "Amino acid", and “Fatty acid metabolism”, were widely
   expressed in lung tumor nodules (Supplementary Fig. [91]2E). Trajectory
   analysis further identified 245 genes (p < 0.001, q < 0.001), including
   35 secretory factors, associated with the transition of Cancer-B1 (Fig.
   [92]2A, B, and Supplementary Fig. [93]2F). KEGG pathway analysis
   indicated the involvement of “Antigen processing and presentation” and
   several metabolic functions, including “Lipoic acid metabolism”,
   “Glycolysis/Gluconeogenesis”, “Carbon metabolism”, “TCA cycle”, "Amino
   acid", and “Fatty acid metabolism” in Cancer-B1 (Fig. [94]2C). Among
   them, Dlat, associated with mitochondrial metabolism, was upregulated
   in all lung cancer nodules in section A and B (Fig. [95]2D, E).
   Moreover, trajectory analysis revealed that Dlat expression gradually
   increased according to the aggressiveness of the primary tumor and was
   higher in all lung metastatic tumor nodules compared with primary
   mammary tumors in the MMTV-PyVT mouse model (Fig. [96]2F–H). The data
   suggested that dysregulation of mitochondrial function induced by Dlat
   may be associated with lung metastasis in TNBC.

Fig. 2. The hub networks contributed to cancers and TMEs.

   [97]Fig. 2
   [98]Open in a new tab

   A The trajectory analysis of lung metastatic cancer nodules in the 4T1
   mouse model. B The soluble factors expressed in different tumor nodules
   of the lung in the 4T1 model. C The KEGG pathway of genes contributing
   to the transition of different tumor nodules in the lung in the 4T1
   model. D The change of Dlat expression in trajectory analysis. E The
   expression of Dlat in different tumor nodules in the lung of the 4T1
   model. F The trajectory analysis of primary and lung metastatic tumor
   nodules in the MMTV-PyVT mouse model. G The change of Dlat expression
   in trajectory analysis and (H) The expression of Dlat in different
   tumor nodules in the primary sites and lung of mice with the MMTV-PyVT
   mouse model.

Upregulation of Dlat was associated with lung metastasis in breast cancer

   We next evaluated the role of Dlat in lung metastatic and primary
   tumors in both models. qRT-PCR revealed that the expression of Dlat
   mRNA in the lung metastatic nodules in the 4T1-L and MMTV cells
   (MMTV-L) isolated from lung tumors compared with corresponding 4T1 and
   primary MMTV cells (MMTV-P) isolated from primary tumors (Supplementary
   Fig. [99]3A, [100]B). Immunohistochemistry (IHC) staining demonstrated
   increased Dlat proteins observed in the lung tumor nodules of mice with
   4T1 breast cancer (Fig. [101]3A). Compared with primary breast cancer,
   elevated Dlat was also found in lung metastatic nodules of MMTV-PyVT
   mice (Fig. [102]3B). The results were consistent with an increased
   expression of Dlat protein in the lung metastatic nodules in the 4T1-L
   and MMTV-L cells (MMTV-L) isolated from lung tumors compared with
   corresponding 4T1 and primary MMTV cells (MMTV-P) isolated from primary
   tumors (Fig. [103]3C).

Fig. 3. Elevated Dlat associated with lung metastasis of TNBC.

   [104]Fig. 3
   [105]Open in a new tab

   The expression of Dlat in the tumor of the primary site and lung
   nodules in the 4T1 model (n = 6) (A) and the MMTV-PyVT model (n = 6)
   (B). C The expression of Dlat in 4T1 cells and MMTV cells. D Inhibition
   of Dlat decreased cell adhesion (24 h) and tumor growth (7 days) in
   lung of mice. Knockout of Dlat reduced cell proliferation of 4T1-L (E)
   and MMTV-L cells (F). Inhibition of Dlat reduced CSC property in 4T1-L
   (G) and MMTV-L cells (H) and, as determined by tumor spheroid formation
   and ALDH activity (I, J). All results were representative of at least
   three independent experiments. Graphs shown as mean ± S.D.
   **p-value < 0.01.

   To validate the role of Dlat in lung metastasis, we established
   Dlat-knockdown or knockout 4T1-L and MMTV-L cells by either shRNA or
   CRISPR/Cas9 system (Supplementary Fig. [106]3D) and then injected them
   into mice via tail vein. The results indicated that the inhibition of
   Dlat decreased cell adhesion (24 h post-injection) and cell
   proliferation (7 days post-injection) in the lungs of mice
   (Supplementary Fig. [107]3E and Fig. [108]3D,). Functional analysis
   revealed that the inhibition of Dlat reduced cell proliferation and
   cancer stem cell properties in 4T1-L and MMTV-L cells (Fig. [109]3E–H,
   Supplementary Fig. [110]3F, [111]H). However, the knockdown of Dlat did
   not affect the ability of cell migration (Supplementary Fig. [112]3I,
   [113]J). These results suggested that Dlat is related to cell adhesion
   and the growth ability of BCs in the pulmonary ecosystem.

Metabolism alternation in Dlat-upregulated lung metastatic nodules

   Dlat has been indicated to be associated with cell metabolism
   [[114]22]. Next, we performed Gene Set Enrichment Analysis (GSEA) to
   explore the biological functions associated with Dlat. Compared with
   Dlat-negative spots in Cancer-B1 (4T1 model) and Cancer-C3 (MMTV
   model), Dlat-positive spots in Cancer-B1 in the 4T1 model and Cancer-C3
   in the MMTV model exhibited prominent enrichment of signatures related
   to “Fatty Acid Metabolism” and “ROS Pathway” (Fig. [115]4A, B,
   p < 0.05). KEGG pathway analysis also showed metabolic pathways related
   to the differentially expressed genes (DEGs) (Top 100) between
   Dlat-negative and Dlat-positive cells in both 4T1 and MMTV mice models
   (Fig. [116]4C, D, Supplementary Tables [117]3 and [118]4).
   Corresponding to the upregulation of lipid utilization genes, higher
   oxygen consumption rate (OCR), basal respiration, ATP-linked
   respiration, maximal respiration, and non-mitochondrial respiration
   were found in both 4T1-L and MMTV-L cells isolated from the lung,
   compared with those isolated from the mammary site of the mice,
   suggesting a greater bioenergetic capacity related to lung metastasis
   (Fig. [119]4E–H). However, although there was an upward trend in the
   extracellular acidification rate (ECAR), it was not statistically
   significant between 4T1 cells isolated from the lung and those from the
   primary site (Supplementary Fig. [120]4). Mitofuel flex assay revealed
   an increased dependence on fatty acid utilization in 4T1-L and MMTV-L
   cells isolated from the lung compared with those isolated from the
   mammary site of the mice (Fig. [121]4I, J). Lower OCR was observed in
   4T1-L cells when Dlat was inhibited by either shRNA transfection or
   CRISPR/Cas9 knockout (Fig. [122]4K, L), suggesting that Dlat was
   associated with regulating metabolism and fuel utility during lung
   metastasis. More importantly, the knockdown of Dlat decreased lung
   metastasis in vivo (Fig. [123]4M), indicating that Dlat-mediated
   metabolism rewriting was required for lung metastasis in TNBC.

Fig. 4. The alternation of lipid metabolism.

   [124]Fig. 4
   [125]Open in a new tab

   GSEA analysis of Dlat-positive and negative spots in the lung
   metastatic nodules in the 4T1 (A) and MMTV-PyVT (B) mouse model. The
   KEGG pathway of DEG between Dlat-positive and negative spots in the 4T1
   (C) and MMTV-PyVT (D) mouse models. The oxygen consumption rate (OCR)
   (E), various parameters of respiration (F) in 4T1 cells. The OCR (G)
   and various respiration parameters (H) in MMTV. Percentage of
   dependency of mitochondrial respiration to oxidize three main energetic
   fuels: glucose, glutamine, and fatty acids in 4T1 (I) and MMTV (J). The
   OCR of Dlat-knockdown (K) and knockout (L) 4T1 cells. M Knockdown of
   Dlat reduced lung metastasis in 4T1 cells in vivo (n = 6). Graphs shown
   as means ± SD of experimental triplicates. *p-value < 0.05;
   **p-value < 0.01; ***p-value < 0.001; ****p-value < 0.0001. DEGs,
   differentially expressed genes.

Increased infiltration of Cd68^+Cd74^+ApoE^+ macrophages of TME surrounding
the metastatic lung nodules

   Metastatic TME is well known to be a critical regulator of cancer fates
   [[126]12], and thus, we further assessed the major cell types in TMEs.
   We sketched the immune profile using the TIMER deconvolution model and
   the suggested different algorithms (CIBERSORT, XCELL) in TIMER2.0. The
   results showed that the scores of activated dendritic cells were
   decreased, whereas M2 macrophages increased in Cancer-B1/2/3-TME and
   diffusive cancer-TME (Fig. [127]5A, B, Supplementary Fig. [128]5A). The
   scores of M2 macrophages and neutrophils increased, particularly in
   TME-B1/2/3 (Fig. [129]5C). However, no differences between all TMEs in
   T cells, B cells, and cancer-associated fibroblasts (CAFs) were
   observed (Supplementary Fig. [130]5B–[131]D). We extracted DEGs from
   TME-B1/2/3, TME-B4, and TME-Diffusive cancer (p < 0.05), and KEGG
   pathway analysis showed that “Antigen processing and presentation” was
   involved in all TMEs, suggesting that antigen-presenting cells (APCs)
   were the major factors for the remodeling of the lung TME for breast
   cancer spreading (Fig. [132]5D, E, and Supplementary Fig. [133]5E).
   Upregulation of 28 genes was found in all TMEs, compared with normal
   lung, and macrophage markers such as Cd74, C1qc, and H2-Eb1 increased
   in all TMEs (Supplementary Fig. [134]5F). The positive correlation of
   these three genes with macrophage markers (Cd68), antigen
   presenting-related factors (H2-Aa, H2-Ab1 H2-Eb1, H2-K, and H2-M3),
   complement system (C1qc), and lysosomal cysteine proteases (Ctss, Ctsb,
   Ctsl) as well as phagocytic receptors (Fcgr1 and Fcgr2) were found in
   all TMEs (Supplementary Fig. [135]5G). Interestingly, the Cd68 surface
   marker and lipid-associated marker ApoE also had a strongly positive
   correlation with Cd74. Most of the spots in TMEs were
   Cd68^+Cd74^+ApoE^+, which had higher expression of M2 type markers than
   Cd68^+Cd74^+ApoE^- spots (Fig. [136]5F). Macrophages isolated from the
   lungs of mice with metastatic 4T1 nodules showed an increased number of
   CD68^+CD74^+ApoE^+ cells, which displayed elevated levels of the M2
   markers CD163 and Arginase I (Fig. [137]5G, H).

Fig. 5. The infiltration of Cd68^+Cd74^+ApoE^+ macrophage in TMEs.

   [138]Fig. 5
   [139]Open in a new tab

   The score of macrophages (A), dendritic cells (B), and Neutrophils (C)
   in TMEs. KEGG pathway of top100 genes of Cancer-B1/2/3-TME (D) and
   Cancer-B4-TME (E). F M2 macrophage score of Cd68^+Cd74^+ApoE^+ and
   Cd68^+Cd74^+ApoE^- spots. G The cell number of Cd68^+Cd74^+ApoE^+
   macrophages in the lungs of mice (n = 10 for control, n = 6 for 4T1
   model). H The expression of M2 markers (Cd163 and Arginase I) in ApoE^+
   and ApoE^-Cd68^+Cd74+ cells isolated from the lung of mice with 4T1
   (n = 6 for 4T1 model). H The expression of AopE in THP-1 after Hs578T
   stimulation. The expression of ApoE (I), Cd206 (J), Cd163 (K), Arginase
   I (L), and IL-10 (M) in macrophage co-cultured with Hs578T cells. All
   results were representative of at least three independent experiments
   in vitro. Graphs shown as mean ± S.D. *p-value < 0.05;
   **p-value < 0.01; ***p-value < 0.001; ****p-value < 0.0001.

   To analyze whether breast cancer cells enhanced ApoE expression in
   macrophages or not and to determine the impact of ApoE on macrophage
   function, we utilized a THP-1/Hs578T TNBC co-culture model and assessed
   the effects of ApoE on the macrophage phenotype by blocking. Figure
   [140]5I shows that Hs578T breast cancer cells increased the expression
   of ApoE (Fig. [141]5I). Furthermore, the expression of M2 markers,
   including CD206, CD163, arginase I, and IL-10, also increased, while
   the M1 markers, CD86 and IL-12, decreased (Supplementary Fig. [142]6A,
   [143]B). Meanwhile, inhibition of ApoE expression in THP-1 cells
   blocked the CD86 downregulation and the CD163 upregulation induced by
   breast cancer cells (Supplementary Fig. [144]6C–[145]E). We also used
   the chemical inhibitor to evaluate the metabolism of breast cancer on
   M2 polarization. The results showed that inhibiting fat utilization can
   effectively reduce M2 macrophage polarization by Hs578T, which was
   supported by findings where a fatty acid inhibitor, etomoxir, prevented
   CD86 downregulation and CD163 upregulation in THP-1 in a co-culture
   system (Supplementary Fig. [146]6F, [147]G). These results suggested
   that macrophages were influenced by breast cancer cells through
   expressing ApoE and shifted toward the M2 phenotype.

Upregulated expression of Lgals1, Thbs1, S100A family in macrophages and
neutrophils of all TMEs

   Next, we applied trajectory analysis to construct the potential
   transition from normal lung niche to TME. Trajectory analysis revealed
   that it began as a normal lung and then progressed to a TME diffusive
   cancer followed by TME-B1/2/3 (Fig. [148]6A). A total of 117 genes
   contributed to the transition from normal lung tissue to TME
   (p-value < 0.05; q-value < 0.05, supplementary Table [149]5).
   Trajectory analysis revealed several soluble factors, including Lagls1,
   Thbs1, Spp1, S100 family (S100A4 and S100A10), Cxcl10, Lmna, and C1q
   (C1qb, C1qc), were associated with all TMEs. Elevated expressions of
   all these factors were also found in all TMEs compared with the
   expressions in normal lungs (Fig. [150]6A). Cxcl16 and Lgals3 were
   related to Cancer-B1/2/3-TME, while Retnla was related to
   Cancer-B4-TME. Col3a1 was upregulated in Diffusive cancer-TME (Fig.
   [151]6B–D). Interestingly, all of these soluble factors were highly
   expressed in the lung metastatic nodules of the MMTV-PyVT model
   (Supplementary Fig. [152]7). Regression analysis showed that Lgals3,
   Lgals1, Spp1, and S100A4, S100A6, had a strong correlation with
   macrophage markers Cd68, Cd74, ApoE, and lysosome enzymes (Fig.
   [153]6E). Multi-immune staining also showed the presence of
   CD68^+ApoE^+Galectin-1^+S100A4^+S100A6^+ macrophages in the metastatic
   nodule of the lung in 4T1 mouse models (Fig. [154]6F). The expression
   of Galectin-1, S100A4, and S100A6 of CD11b^+CD68^+ApoE^+ macrophages
   was higher compared with the CD11b^+CD68^+ApoE^- macrophages, as
   determined by the fluorescence intensity (Supplementary Fig. [155]8).

Fig. 6. The factors contributed to the remodeling of TMEs.

   [156]Fig. 6
   [157]Open in a new tab

   A Trajectory analysis of all TMEs. The genes associated with
   Cancer-B1/2/3-TMEs (B), Cancer-B4-TME (C), and Diffusive cancer-TMEs
   (D). E The correlation of various soluble factors with macrophage
   markers. F The presence of CD68^+ApoE^+S100A4^+S100A6^+Galectin-1^+
   macrophages in lung TME in vivo (n = 6). G, H The interaction of
   Cancer-B1/2/3 with their TMEs. I, J The interactions between
   Cancer-B1/2/3-TMEs with Cancer-B1/2. K The interactions within
   Cancer-B1/2/3-TMEs.

   To clarify the interaction networks in the TMEs and lung metastatic
   cancer, we performed ligand-receptor pair analysis on the ST
   transcriptomic data of all TMs in the 4T1 mouse model. The interactomes
   revealed a putative interaction between Lgals3 on neoplastic epithelial
   cells and its receptor on cell components of Cancer-B1/2/3-TME,
   respectively (Fig. [158]6G, H). In addition, the cells of Cancer-B1/2/3
   secreted Thbs1 to interact with Cancer1/2 by Cd36 and Itga2 (Fig.
   [159]6I, J). S100A8 and S100A9/Cd93 were major communication axes
   within Cancer-B1/2/3-TME (Fig. [160]6K).

Exploration of spatial proteasome in TME and lung metastatic BC tumor

   Finally, we utilized multiplex immunofluorescence staining, CODEX, to
   analyze the spatial proteome. Through staining with 21 antibodies
   determined by ST data, cells were classified into 28 clusters (Fig.
   [161]7A). To eliminate bias introduced by the distance between
   interacting cells, we conducted spatial analysis calculations for
   cell-cell communication. The results showed that 19 cell clusters,
   including cancer cells (3 subsets, Ki-67^+-, Ki67^+, and Ki67^-), four
   groups of macrophages, one group of monocytes, one group of Ly6g^+
   cells (neutrophils), and six groups of endothelial cells (CD31^+
   cells), as well as four immune cell populations, exhibited cell-cell
   interactions (Fig. [162]7B and sTable [163]5). Cluster 25 cells
   (CD163^+ApoE^+ M2 macrophages) and Cluster 18 (cancer cells) showed the
   most active cellular interactions (Fig. [164]7C). Cluster 25
   (CD163^+ApoE^+) macrophages expressed S100A4 and S100A10 (Fig.
   [165]7D), while Cluster 18 (ApoE^+ki67^+) cancer cells expressed high
   levels of ApoE, Ki-67, and vimentin (Fig. [166]7E). Cluster 23
   (CD31^+ApoE^+ EC cells) also exhibited a significant spatial
   interaction with Cluster 18 cancer cells by S100A10 and Thbs1 secretion
   (Fig. [167]7F). The image data also revealed that Cluster 25/18 or
   Cluster 23/18 were located in close proximity. Interestingly, we
   discovered an intriguing subset of proliferative neutrophils
   (Ly6g^+ApoE^+Ki67^+) that express cell proliferation markers (Ki67) and
   S100A4 and Thbs1. This specific neutrophil subset exhibited spatial
   interactions with ApoE^+ cancer (Cluster 18) (Fig. [168]7G).
   Furthermore, upon isolating macrophages and Ly6g^+ neutrophils from
   lungs with 4T1 breast cancer metastasis, we found that macrophages
   expressed higher levels of ApoE, S100A10, and S100A4 than those
   macrophages isolated from normal mice (Fig. [169]7H). Additionally,
   neutrophils isolated from lungs within metastatic 4T1 tumor cells show
   higher expression of ApoE and S100A4 compared with neutrophils from the
   lungs of mice without tumors (Fig. [170]7I).

Fig. 7. Spatial proteomics of lung metastatic niche.

   [171]Fig. 7
   [172]Open in a new tab

   A The cell clusters of lung TME were conducted using CODEX staining. B
   The spatial analysis of cell clusters in lung TME. C The odd ratio of
   cell communication between different cell clusters. D CD163^+ApoE^+
   macrophages expressed S100A10 and S100A4 protein. E ApoE^+Ki67^+ cancer
   cells expressed higher levels of vimentin. F CD31^+ApoE^+ endothelial
   cells expressed S100A10 and Thbs1 protein. G The expression of S100A4
   and Thbs1 protein in Ly6g^+ neutrophils. H The levels of ApoE, S100A4,
   and S100A10 in the macrophages isolated from the lung with or without
   4T1 tumor nodules (n = 6). I The levels of ApoE and S100A4 in the
   neutrophils isolated from the lung with or without 4T1 tumor nodules.
   Graphs are shown as means ± SD of experimental triplicates.
   *p-value < 0.05; **p-value < 0.01; ***p-value < 0.001;
   ****p-value < 0.0001.

Discussion

   TNBCs are highly aggressive breast cancers (BCs) with clinical features
   of high metastatic dissemination, susceptibility to relapse, and poor
   overall survival [[173]23]. Over the last few years, technological
   advancements in single-cell profiling and spatial mapping have
   revolutionized the understanding of tumor heterogeneity and tumor TMEs
   in primary sites of TNBCs [[174]24, [175]25]. However, the significance
   of the spatial organization and tumor heterogeneity of TNBC in the lung
   has not been investigated. Here, we aimed to characterize both the
   tumor heterogeneity of lung metastatic nodules and the influence on the
   tumorigenic TME in the lung. We uncovered insights into the requirement
   for lipid metabolism rewriting for adhesion and sustained growth during
   lung metastasis in TNBC. In addition, cell components of the TME,
   including macrophages, neutrophils, and endothelial cells, contributed
   to the remodeling of the TME and interactions with cancer cells when
   they expressed higher levels of lipid metabolism protein, ApoE. Our
   findings indicate that changes in lipid metabolism play a pivotal role
   in influencing lung metastasis in both breast cancer cells per se and
   the cells within the lung metastatic niche.

   Currently, the metabolic adaptation of organ-specific metastasis has
   garnered attention and is considered a crucial factor in enabling tumor
   cells to successfully establish themselves in a new metastatic
   microenvironment [[176]26]. In the oxygen-rich environment of the
   lungs, metastatic cancer cells adapt by shifting their metabolism from
   glycolysis to oxidative phosphorylation (OXPHOS). Studies have shown
   that increased lipid utilization plays a pivotal role in driving lung
   metastasis of breast cancer [[177]27, [178]28]. A recent study also
   demonstrated that the lung is a lipid-rich microenvironment, and
   targeting the process of palmitate inhibits lung metastasis of breast
   cancer [[179]29]. In contrast, the degradation of lipids by fatty acid
   amide hydrolase (FAAH) prevents tumor progression and lung metastasis
   both in vitro and in vivo, and FAAH serves as a biomarker for longer
   survival in breast cancer patients [[180]30]. Dlat is a component of
   the pyruvate dehydrogenase (PDH) complex within the TCA cycle and has
   been reported to promote carcinogenesis by increasing glycolysis
   [[181]31]. Upregulated Dlat has been reported to be associated with
   poor clinical outcomes in various cancers, including pancreatic
   adenocarcinoma, liver cancer, and breast cancer via using
   bioinformatics analysis [[182]32–[183]34]. In our study, we identified
   that Dlat and lipid in lung metastatic breast cancer were increased.
   Notably, increased OCR, ECAR, and lipid dependence were observed in 4T1
   and MMTV cells isolated from metastatic lung tumor nodules of mice.
   Inhibition of Dlat reduced the OCR in TNBC cells, indicating that Dlat
   upregulation is crucial for metabolic reprogramming during lung
   metastasis. Dlat knockdown/knockout also diminished the proliferative
   capacity and cancer stem cell properties of TNBC cells. Also,
   loss-of-function experiments revealed that Dlat knockdown decreased
   cell adhesion and inhibited lung tumor growth in vivo. Altogether,
   these findings provide compelling evidence that alterations in fuel
   utilization contribute to the pathogenesis of triple-negative breast
   cancer, particularly lung metastasis, potentially through the
   modulation of Dlat expression.

   Apolipoprotein E (ApoE) has been indicated to be expressed
   predominantly by tumor-associated macrophages (TAM) or
   cancer-associated fibroblasts in the tumor in mice or humans
   [[184]35–[185]37]. Cancer in ApoE knockout (ApoE^−/−) mice has a lower
   tumor burden, less fibrosis, reduced innate immune response, and
   increased adaptive immune cell infiltration while inhibiting ApoE
   potentiates immune checkpoint inhibitors (ICIs) effectiveness for
   cancer [[186]36]. Complement C1q (C1q) was also considered to act as a
   marker of tolerogenic and immunosuppressive macrophage populations in
   both healthy and tumor tissues. C1q is co-expressed in TAMs with human
   leukocyte antigen DR (HLA-DR), ApoE, and C1q expression in TAM
   associates with T cell exhaustion, ICI non-responders, and poor
   prognosis in patients with malignancy [[187]35, [188]38, [189]39].
   Huggins et al. reported that lipid-associated TAMs (LAMs,
   CD68^+Lagls3^+Trem2^+Fabp5^+), which were observed in the lung of 4T1
   nodules, are related to lipid metabolism, extracellular matrix
   remodeling, and immunosuppression [[190]40]. In our study, we revealed
   that Cd68^+ApoE^+ cells were found in the TMEs of lung metastatic
   breast tumors were correlated with C1qa/b/c, Cd74, and ApoE expression.
   Moreover, Cd68^+Cd74^+ApoE^+C1qa^+ spots exhibited an M2 phenotype and
   expressed higher levels of galectin-1 and S100A4/6, resulting in
   immunosuppression in the lung. Our results indicate that
   Cd68^+Cd74^+ApoE^+ macrophages play a crucial role in breast cancer
   metastasis to the lungs. However, our findings revealed the
   upregulation of 28 genes within the lung metastatic TME. Further
   investigation is required to identify the specific cellular
   subpopulations in which these genes are expressed.

   ApoE is produced and released by various cell types, including
   astrocytes, microglia, vascular mural cells, hepatocytes, and
   macrophages [[191]41], and it may play different roles in different
   cell types. ApoE4 expression in vascular mural cells correlated with
   the activation of astrocytes, whereas ApoE3 was associated with an
   angiogenic signature in the cerebrovascular system [[192]41]. ApoE has
   been reported to play a role in the development of breast cancer
   [[193]42]. Additionally, in prostate cancer, it has been observed that
   the secretion of ApoE can regulate the phenotype of neutrophils,
   promoting an immunosuppressive function [[194]43]. Our results reveal a
   subset of tumor cells expressing high levels of ApoE, with
   ApoE-positive endothelial cells (ECs) and ApoE-positive neutrophils
   positioned close to each other.

   Moreover, these ApoE-rich ECs and neutrophils secrete elevated levels
   of inflammatory substances, including S100A4, S100A10, and Thbs-1.
   Consequently, these ApoE-rich ECs and neutrophils actively participate
   in remodeling an immunosuppressive metastatic niche. However, the
   mechanisms underlying the delivery of ApoE between tumor cells-ECs or
   tumor cells-neutrophils and the regulation of inflammatory substances
   in individual cell types by ApoE upregulation, as well as the
   contribution of all secretory factors in the TME remodeling require
   further investigation.

   In conclusion, this study enhances our understanding of TNBC
   heterogeneity and maps its microenvironmental landscape using spatial
   omics approaches. Our findings indicate that the shift from glucose
   dependence to fatty acids is a critical change for lung metastasis in
   the TNBC, and the upregulated expression of Dlat plays a vital role in
   driving this process. Nevertheless, our data depict a detailed
   landscape of both the most common and the most aggressive tumor
   microenvironment, untangling the effects of the niche on cancer fate in
   the lung through various soluble factor-mediated interactions. This
   study delivers valuable data that serve as an invaluable resource for
   further research, delineating spatial genomic and proteomic features
   that can drive the development of therapeutic strategies to improve the
   clinical outcomes of patients with TNBC.

Supplementary information

   [195]Suppementary file^ (6.1MB, pdf)
   [196]Western blot^ (373.5KB, pdf)
   [197]Suppementary Table 3^ (13.6KB, xlsx)
   [198]Suppementary Table 4^ (225.5KB, xlsx)
   [199]Suppementary Table 5^ (25.5KB, xlsx)
   [200]Suppementary Table 6^ (18.4KB, docx)

Acknowledgements