Main reagents
Paclitaxel (P815862, MACKLIN) was dissolved in a mixture of corn oil and ethanol. Human TNFα (AF-300-01A), murine TNFα (AF-315-01A) and murine IFNγ (AF-315-05) were purchased from Peprotech. Anti-TNFR2 antibody (TR75-54.7, Bio X Cell) was used for the in vivo experiments. Anti-TNFR1 (113103, BioLegend) and TNFR2 (113305, BioLegend) antibodies were prepared for the in vitro experiments. SC75741 (S7273, Selleck) is a potent inhibitor for NF-κB activation.
Cell and culture conditions
sEND.1 cell (RRID: CVCL_6270) is a sort of immortalized vascular endothelial cell from skin angioma and was established in Dr. Blankenstein’s laboratory [21]. Human umbilical vein endothelial cell (HUVEC; RRID: CVCL_2959) and the mouse triple-negative breast cancer cell line 4T1 (RRID: CVCL_0125) were purchased from ATCC (Manassas, VA). These cells were cultured in mycoplasma-free Dulbecco’s modified Eagle’s medium (DMEM; SH10022.01, Hyclone), supplemented with 10% FBS (P30-3302, PAN) and 1% penicillin/streptomycin (SV30010, Hyclone).
Tumor model and treatment
Sex-matched (female) and age-matched (6–8 weeks old) BABL/c wild-type (WT) mice were purchased from Vital River Laboratories (Beijing, China) and randomized to the different experimental groups. All the mice were maintained under specific pathogen-free conditions. 4T1 cells (0.5 × 106/per) in 100 μl PBS were subcutaneously (s.c.) injected into breast pad of mice. Tumor growth was monitored every 2–3 d during the entire experimental period. Tumor volumes (V) were assessed as follows: V = 0.5 × (L × W)2. L represents the long diameter of the tumor mass and W represents the short diameter. Starting on day 5–7 post inoculation, paclitaxel (20 mg/kg) agent, Anti-TNFR2 (10 mg/kg) antibodies, or PBS as a control were intraperitoneally (i.p.) injected into the mice when the tumor volume reached approximately 100 mm3. Additional treatment scheduling is shown in Figs. 2A and 8A.
RNA-sequencing analysis
The total RNA of sEND.1 cells administrated with co-stimulation of TNF (20 ng/ml) and IFNγ (10 ng/ml), or IFNγ (10 ng/ml) as control for 24 h, was extracted by mRNA isolation kit V2 (L/N 7E522J1, Vazyme, China) and then RNA sequencing was performed by BGI Genomics (Shenzhen, China). Differentially expressed genes (DEGs) were defined at the settled condition of P_adjust < 0.05 and a log2 fold change (FC) > 0.5. Subsequent data from RNA-sequencing (RNA-seq), such as volcano plots, heatmaps, Kyoto Encyclopedia of Genes and Genomes (KEGG) and Gene set enrichment analysis (GSEA) were obtained from Dr. Tom’s website created by BGI.
Flow cytometry
The expression of PD-L1 or TNFR2 on EC was detected using flow cytometry (FCM). These single-cell suspensions were prepared in vitro or from tumor tissues and then stained with antibodies specific for CD45− Percp-cy5.5 (157612, BioLegend), CD31-FITC (160212, BioLegend), PD-L1-PE (12-5982-82, eBioscience), mouse TNFR2-PE (550086, BD) and human TNFR2-PE (552418, BD). Similarly, suspensions for CD8+ T cells isolated from tumor tissues were stained with antibodies specific for CD45-Percp-cy5.5, CD3-FITC (100203, BioLegend), CD8-APC (100712, BioLegend) and PD-1-PE (135205, BioLegend). Cell suspensions were performed using FACS Calibur devic (BD, USA) and raw data was analyzed using Flow Jo software (RRID:SCR_008520; BD, USA).
Western blotting
sEND.1 cells were harvested and lysed with RIPA buffer (R0020, Solarbio), supplemented with PMSF (P8340-1, Solarbio) and a protease inhibitor cocktail (P8340, Sigma). Cell lysates were collected in RIPA buffer and the protein concentration was quantitated by Protein BCA Assay Kits (23228, Thermo Fisher Scientific). An equal number of proteins was separated by SDS-PAGE and transferred onto nitrocellulose membranes, followed overnight incubation with rabbit anti-NF-κB antibody (65 kDa, 1:1000, 8242 s, CST), rabbit anti-pNF-κB antibody (65 kDa, 1:1000, 3033 s, CST), rabbit-anti-Glut1 (54 kDa, 1:1000, 07-1401, Millipore), rabbit-anti-HK2 (102 kDa, 1:1000, 2867 s, CST), rabbit-anti-pPFKFB2 (55 kDa, 1:1000, 13064 s, CST), mouse anti-β-actin antibody (42 kDa, 1:5000, AC004, abclonal). The nitrocellulose membranes were then incubated with horseradish peroxidase (HRP)-conjugated secondary antibodies. Protein bands were visualized using an eECL Western Blot Kit (CW00495, CwBio) and detected using a ChemiDoc MP Imaging System (Bio-Rad).
Quantitative real-time PCR (qRT-PCR)
The total RNA was extracted from sEND.1 cells using mRNA isolation kit V2 (L/N 7E522J1, Vazyme) according to the manufacturer’s instructions, then converted into cDNA using the Prime Script™ RT reagent kit (RR047A, Takara) and subjected to subsequent qPCR using SYBR Premix Ex Taq II (RR820A, Takara). The primers for mouse mRNAs are listed as follows: Gapdh forward: TCTCTGCTCCTCCCTGTTCC, reverse: TACGGCCAAATCCGTTCACA; Pd-l1 forward: TGCGGACTACAAGCGAATCACG, reverse: CTCAGCTTCTGGATAACCCTCG; Pvr forward: GAGGCAGTAGAAGCACCAATGC, reverse: GGTGACCATTGGCAGAGATGCA; Vista forward: AACAACGGTTCTACGGGTCC, reverse: CGTGATGCTGTCACTGTCCT; Tigit forward: CCACAGCAGGCACGATAGATA, reverse: CATGCCACCCCAGGTCAAC; H2-d1 forward: TGAGGAACCTGCTCGGCTACTA, reverse: GGTCTTCGTTCAGGGCGATGTA; H2-k1 forward: GGCAATGAGCAGAGTTTCCGAG, reverse: CCACTTCACAGCCAGAGATCAC; H2-q1 forward: GCTGTTCTGGTTGTCCTTGGAG, reverse: AGAGCACAGTCCTCTCCTTGTC.
Immunofluorescence staining
Tumor tissues were isolated and embedded in cutting temperature compound (OCT) and therewith prepared as cryostat sections. Frozen tissue sections (7 μm) were fixed with 4% paraformaldehyde for 15 min and permeabilized three times at room temperature. After blockade by 2% bovine serum albumin, the sections were incubated overnight with the following primary specific antibodies: rat anti-CD31 (1:200, 550274, BD), rabbit anti-TNFR2 (1:200, ab109322, Abcam) and rabbit anti-CD8 (1:400, ab217344, Abcam). Following incubation with fluorescence-conjugated secondary antibodies (1:400), the tissue sections were stained with DAPI (1:3000). Images were acquired using an inverted fluorescence microscope (Leica).
Manders’ colocalization coefficients
To better quantify the fraction of one protein that colocalizes with another distinct protein, Manders’ Colocalization Coefficients (MCC) were created by Manders et al. [22]. In this study, we determined and quantified the dynamics of spatial relationship between ECs and CD8+ T cells using MCC analysis. M1 in an image represents the ratio of the part of CD8 protein that colocalizes with another protein (CD31) to the whole, that is, the overlapping fraction of CD8; M2 represents the same pattern of CD31. Biological image analysis of MCC performed using an auto plug-in in Fiji/Image J (RRID: SCR_003070; NIH). The formulae for calculating value of M1 and M2 in this study are as follows:
$${M_{1}} = \frac{{\sum\nolimits_{i} {{CD8}_{i,colocal} } }}{{\sum\nolimits_{i} {{CD8}_{i} } }}\quad {M_{2}} = \frac{{\sum\nolimits_{i} {{CD31}_{i,colocal} } }}{{\sum\nolimits_{i} {{CD31}_{i} } }}$$
Glucose analog uptake
As a glucalogue, 2-[N-(7-Nitrobenz-2-oxa-1,3-diaxol-4-yl) amino]-2-deoxyglucose (2-NBDG) can be taken up by cells through glucose transmembrane transfer. sEND.1 cells (1 × 105 cells) were seeded in 24-well plates. After treatment with or without TNFα for 6 h, cells were rewashed with PBS, followed by an addition of low glucose culture media supplemented with 100 µM 2-NBDG (N13195, Life Technologies) and incubation for 45 min at 37 °C. An additional group lacking 2-NBDG was set as blank control. Next, the cells were harvested and 2-NBDG uptake was immediately detected by flow cytometry in the distinct groups.
Cellular glycolysis assay
The change in extracellular acidification rate (ECAR) for ECs regulated by TNFα was detected. sEND.1 cells were seeded into seahorse XF-96 microplate (6000 cells per well) for overnight, and then treated with or without TNFα (20 ng/ml) for 6 h. Culture medium was replaced with the XF base medium (pH 7.4) supplemented with 2 mM l-glutamine (V900419, Sigma-Aldrich), followed by incubated in a non-CO2 incubator for 30 min. Subsequently, glucose, oligomycin, and 2-deoxy-d-glucose (2-DG) were added into the medium at final concentration of 10 mM, 1 μM, 50 mM, respectively. Finally, ECAR was measured by Seahorse XF-96 extracellular flux analyzer (Seahorse Bioscience, Agilent) and the corresponding number of ECs was normalized by hoechst 33342 (C0030, Solarbio) according to the manufacturer’s instructions.
CD8+ T cell proliferation assays
After filtration of grinded spleens freshly isolated from tumor-free OT-I mice (003831, Jackson Laboratory), splenocytes were incubated overnight in RPMI1640 complete medium containing IL-2 (10 ng/ml; AF-212-12, Peprotech) to remove adherable cells. The next day, splenocytes were labeled with CFSE (5 mM; 21888, Sigma-Aldrich) for 10 min and then neutralized with RPMI1640. Both splenocytes (30 × 104) and sEND.1 (6 × 102) were pre-treated by IFNγ (20 ng/ml) and TNFα (10 ng/ml). Anti-TNFR2 antibody (10 μg/ml) was pre-treated 1 h prior to administration of TNFα. The cell mixture in a total of 100 µl culture medium containing IL-2 (10 ng/ml), SIINFEKL (0.1 μM; S7951, Sigma-Aldrich) and anti-mouse CD3 (3 μg/ml; 550275, BD Biosciences) and CD28 antibodies (1 μg/ml; 557393, BD Biosciences) was dispensed into a well of 96-well round-bottom plates and cultured for at least 72 h at 37 °C. CD8+ T cells were sorted using specific CD45, CD3 and CD8 staining. The dilution of CFSE on CD8+ T cells was determined by FCM.
ScRNA-seq data acquisition
Single-cell RNA sequencing (scRNA-seq) raw data, comprising unique molecular identifier (UMI) matrix were downloaded from Gene Expression Omnibus (GEO) with accession number GSE169246. As previously described [23], 10 treatment-naïve patients diagnosed with advanced TNBC who received paclitaxel monotherapy based on clinical routine as first-line treatment. According to the Response Evaluation Criteria in Solid Tumors (RECIST), these patients were divided into responders (PR; P022, P011, P020, P008, and P013) and non-responders (SD/PD; P025, P018, P023, P024, and P003). Matched tumor biopsy and peripheral blood samples were collected before and after chemotherapy for scRNA-seq sequencing.
Data reprocessing for dimension reduction and unsupervised clustering
R v4.0 (Seurat) was applied to analyze the raw UMI matrix through reserving high-quality cells with mitochondrial gene counts less than 20% and filtering out genes detected in less than five cells and cells with fewer than five genes. Next, based on the raw UMI counts, the normalization of matrix data following the normalization of total counts per cell was calculated, scaled by 1e6 and logarithmically transformed. The dispersion-based method, Seurat (FindVariableFeature) was used to calculate the highly variable genes (HVGs) and selected the top 2000 HVGs, further regressing out the unwanted variation, such as total counts and outliers of mitochondrial gene proportion from normalized matrix data. Principal component analysis (PCA) for dimension reduction was performed, and the top 20 primary components were selected for subsequent unsupervised clustering. Based on PCA, the Louvain algorithm was used to construct the SNN graph and determine clustering at a resolution of 0.6 for the first-round (re-clustering for individual cell types at a resolution of 0.4). Uniform Manifold Approximation and Projection (UMAP) or t-Distributed Stochastic Neighbor Embedding (t-SNE) projections were performed for cluster visualization.
Identification for cell type
Cell-type markers were obtained from the CellMarker 2.0 database (http://bio-bigdata.hrbmu.edu.cn/CellMarker/) for identification. Each cluster was manually assigned cell types in terms of normalized gene expression using the following canonical markers: endothelial cells (PECAM1, CLDN5), T cells (CD3D, CD3E, CD3G, TRBC1), myeloid cells (LYZ, MNDA) and cancer cells (FAM177A1, B3GNT2, TRPS1). Further, we re-clustered endothelial cell and T cell types individually. Each EC subcluster was also defined by the top 100 markers (Additional file 2: Table S1), such as EC_subcluster2 (JUNB, IFITM1, HLA-C, HLA-DPA1) and EC_subcluster4 (PTPRC, MALAT1, ITGA4). Likewise, T cells were re-clustered at same resolution matched with ECs. T lineage subcluster identification was determined by the top 100 dominant markers, for example, exhausted T_subcluster0 (CD8, LAG3, NR4A2, and CXCL13) and Treg_subcluster1 (CD4, TNFRSF18, TNFRSF4, and GK).
DEGs and functional enrichment analysis
In this study, the DEGs between pre- and post-treatment from responders/non-responders were screened by the Seurat (FindAllMarkers) function (|log2FC| > 0.5, adjusted P-value < 0.05). To compare the changes in the immune pathways of EC or T cell subtypes under different conditions, the enrichment score of genes involved in immune pathways was obtained using escape package (ssGSEA). For revealing biologic nature of cell subclusters, GSEA was performed by the clusterProfiler and similarly, Gene Ontology (GO) enrichment analysis was performed by clusterProfiler (bubble plot) and GOplot (circular plot) package.
Correlations analysis
To reveal the correlation between different molecules, according to the average expression of genes in cell subclusters from the scRNA-seq dataset, Pearson’s correlation was performed using the R software. A customizable functional module for correlation analysis on the GEPIA2 website (http://gepia2.cancer-pku.cn) was investigated. To further elaborate on the relationship of TNFR2 with ECs or subtypes of ECs, according to the specific marker genes of cell populations, the score of each cell of ECs or EC_sub2 was calculated by escape package (ssGSEA). Through regression with a linear module between the total score of cell population and TNFR2 expression, P-value and co-efficient of determination R2 (R-squre) were obtained.
Spatial transcriptomics data acquisition and spatial visualization
The raw unique molecular identifier (UMI) matrix of spatial transcriptomics (ST) data was obtained from GEO via the accessible number GSE210616, including 43 samples from 22 patients with primary TNBC [24]. A procedure similar to scRNA-seq data reprocessing, like normalization of gene expression counts, technical bias correction, further dimensionality reduction and clustering, and spatial visualization of either cell types or molecules at each image site was performed by R software according to positive definition for cell types or gene expression levels, respectively. The score of individual cell type were calculated by ssGSEA based on cell type markers from CellMarker 2.0. The portion of cells located in regions with a score greater than the top 25% were defined as positive expression; otherwise, it was defined as negative. Of note, any imaging site could concurrently show positivity for multiple cell types.
Join count analysis
For the sake of quantifying spatial correlation between different cell types, join count analysis (JCA) developed by Bassiouni et al. [24] was introduced in this study and stJoincount (https://github.com/Nina-Song/stJoincount) was used for quantitation of spatial relationships between two types of cells. Briefly, a unique coordinate position for each image site in the samples was set using getcoord function of stJoincount. Spatial dependencies were quantified for each cell-type pair. We assigned a cell-type pair to the image site, even if the cell types were both scored in a shared site, by randomly assigning one cell type to image sites with the coordinate (x, y) value (x + y) being even, and another cell type to those with the coordinate (x, y) value (x + y) being odd (Additional file 1: Fig. S3F). The spatial distance between cell types was then determined quantitatively using the stJoincount package as previously described [24].
Human protein atlas
Immunohistochemical images of TNFR2 proteins in human breast cancer tissues were obtained from the Human Protein Atlas (HPA) website (https://www.proteinatlas.org/ENSG00000028137-TNFRSF1B) to detect their distribution around the vessel regions.
Protein–protein interaction networks
The database of STRING (https://cn.string-db.org/cgi/input.pl) was searched to explore the protein interactions of “TNFR2” with immune checkpoints involved in T cell-regulatory pathways through publicly integrated both experimental interaction results and computational interaction predictions. In the “Multiple Proteins” channel, TNFR2 and molecules involved in pathway “negative regulation of activated T cell proliferation” was inputted.
Statistical analysis
All data were analyzed using GraphPad Prism software (RRID: SCR_002798). Bioinformatic plots and statistical analyses were carried out using R software (version 4.0; RRID: SCR_001905). Comparisons were performed using unpaired Student’s t-test, Two-way ANOVA, or Friedman test. Data were expressed as the mean ± SD. Statistical significance was set at P-value < 0.05.
Add Comment