Inflammatory breast cancer (IBC) is a highly aggressive subtype of breast cancer characterized by rapidly arising diffuse erythema and edema. Genomic studies have not identified consistent alterations and mechanisms that differentiate IBC from non-IBC tumors, suggesting that the microenvironment could be a potential driver of IBC phenotypes. Here, using single-cell RNA sequencing, multiplex staining, and serum analysis in patients with IBC, we identified enrichment of a subgroup of luminal progenitor (LP) cells containing high expression of the neurotropic cytokine pleiotrophin (PTN) in IBC tumors. PTN secreted by the LP cells promoted angiogenesis by directly interacting with the NRP1 receptor on endothelial tip cells located in both IBC tumors and the affected skin. NRP1 activation in tip cells led to recruitment of immature perivascular cells in the affected skin of IBC, which are correlated with increased angiogenesis and IBC metastasis. Together, these findings reveal a role for cross-talk between LPs, endothelial tip cells, and immature perivascular cells via PTN–NRP1 axis in the pathogenesis of IBC, which could lead to improved strategies for treating IBC.

Significance:

Nonmalignant luminal progenitor cells expressing pleiotrophin promote angiogenesis by activating NRP1 and induce a prometastatic tumor microenvironment in inflammatory breast cancer, providing potential therapeutic targets for this aggressive breast cancer subtype.

Inflammatory breast cancer (IBC) is a rare and highly malignant subtype of breast cancer, and approximately 20% to 30% of patients with IBC present with de novo distant metastasis and a 5-year overall survival rate of only 40% to 60% (1, 2). The distinctive clinical characteristics of IBC include rapid onset of skin edema and erythema, occupying at least one-third of the breast, which serve as diagnostic criteria and are thought to be defining characteristics of its highly aggressive nature (3). However, the underlying mechanisms that cause inflammation and promote metastasis have not been elucidated. IBC shares the same molecular subtypes as non-IBC, even though they affect different proportions of patients: hormone receptor-positive (HR; 30% in IBC vs. 60%–80% in non-IBC), HER2-positive (40% in IBC vs. 25% in non-IBC), and triple-negative (30% in IBC vs. 10%–15% in non-IBC; refs. 1, 4, 5). Despite the histopathologic similarity between IBC and non-IBC, the malignant cells of IBC are usually dispersed in clusters throughout the breast and affected skin, leading to florid tumor emboli (6, 7). Moreover, there are currently no IBC-specific therapeutic strategies available (8). Therefore, it is urgent to explore the pathogenesis and molecular mechanisms that lead to the development of IBC to identify novel targets to guide treatment.

Genomic studies have not yet identified consistent specific gene signatures and biological mechanisms that distinguish IBC from non-IBC tumors (4, 9, 10). Tumors have complex ecosystems that are defined by spatiotemporal interactions among heterogeneous cell types (11, 12). A preclinical study revealed that mesenchymal stem cells significantly increased the clinical features of skin invasion and metastasis in an IBC cell line SUM149-based xenograft model (13), highlighting the microenvironment as a potential driver of IBC phenotypes. In addition, mammary stem cells are also enriched in adjacent normal IBC tissues (14), which indicates that the mammary epithelium might not be an innocent bystander. The normal mammary epithelium is composed of two primary cell lineages that are arranged in a bilayer structure (15). The luminal layer lining the ducts and the alveoli is composed of luminal progenitor (LP) cells that exhibit progenitor-like characteristics and luminal mature cells that are involved in hormone sensing and milk production. The contractile myoepithelial layer with a basal location is composed of basal cells and myoepithelial cells that function in maintaining the integrity of ducts and alveoli. However, few studies have comprehensively and deeply dissected the cellular heterogeneity and phenotypic characteristics of IBC to date. These findings and the challenge in identifying drivers of IBC suggest that exploring the cellular composition and uncovering key molecular mechanisms in the microenvironment may reveal the unique biological behavior of IBC.

In this study, we collected paired tumor and skin specimens from patients with IBC and non-IBC prospectively for single-cell RNA sequencing (scRNA-seq), bulk RNA sequencing (RNA-seq), and whole-exome sequencing (WES). We found that a subpopulation of LP cells secreting pleiotrophin (PTN) was significantly enriched in patients with IBC. The secreted PTN then interacts with neuropilin 1 (NRP1) on endothelial cells (EC) to stimulate tip cell sprouting, resulting in the accumulation of immature perivascular (PVL) cells in the affected skin of patients with IBC. Finally, these immature PVL cells contributed to the aggressive nature and metastatic potential of IBC. Thus, our data offer new insight into the distinct biology of IBC and provide a valuable resource for the development of novel therapeutic approaches to target IBC.

Patient samples

Patient samples were obtained from the First Affiliated Hospital of Sun Yat-sen University, Traditional Chinese Medicine Hospital of Guangdong Province, Jiangmen Central Hospital, Maternal and Child Health Care Hospital of Guangdong Province, Jieyang People's Hospital, and Sun Yat-sen University Cancer Center.

A total of 94 patients with IBC and 228 patients with non-IBC were enrolled in this study. Paired, fresh tumor and skin tissue samples were collected from the discovery cohort, which consisted of treated naïve 17 patients with IBC and 5 patients with non-IBC, and detailed clinical and pathologic information, including age, TNM stage, and molecular subtype, is shown in Supplementary Table S1. Formalin-fixed paraffin-embedded (FFPE) tissue blocks of tumors, adjacent normal tissues, skin, and serum were collected from 52 patients with IBC and 69 patients with non-IBC from the validation cohort. The available clinical features of the patients in the validation cohorts are summarized in Supplementary Table S2. We obtained scRNA-seq data from published studies, including 66 tumor samples from non-IBC patients and 11 normal breast gland samples from healthy women. We used abbreviations for the names of each center to name each cohort, including non-IBC_Tumor-KULEUVEN, non-IBC_Tumor-TKCC, non-IBC_Tumor-WEHI, and Normal_breast-WEHI. This information is summarized in Supplementary Tables S3 and S4. In addition, we downloaded RNA microarray data (GSE22597) of 25 IBC and 57 non-IBC tumors from the MD Anderson Cancer Center, which were named IBC_Tumor-MDA. Furthermore, we included the exome sequencing data from 33 patients with non-IBC, which were downloaded from The Cancer Genome Atlas (TCGA) database, in our analysis.

scRNA-seq data generation

Biopsies were obtained via diagnostic needle biopsy with a 14-G needle, immersed in complete medium containing tissue storage solution (130–100–008; Miltenyi Biotec), and transported to the laboratory in a refrigerated container. Suitable small tissue blocks (no adipose tissue) were cut into pieces (diameter 1–2 mm). The tumor pieces were then transferred to a tube containing 5 mL of dissociation enzyme mixture (collagenase I + collagenase II + neutral protease + hyaluronidase); meanwhile, the skin pieces were then transferred to a tube containing 5 mL of tissue dissociation enzyme mixture (collagenase IV + DNase I + neutral protease + hyaluronidase) and incubated at 37°C for 30 minutes on a shaker. The digested mixture was passed through a 70 μm cell strainer (130–098–462; Miltenyi Biotec) and treated with 1× RBC lysis buffer (00–4333–57; eBioscience). Single cells were processed with the GemCode Single-Cell Platform using Gemcode Gel Bead, Chip, and Library Kits (10× Genomics) following the manufacturer's protocol with modifications. An appropriate volume of cell suspension with approximately 20,000 cells was loaded in each channel, resulting in the capture of 8,000 cells per sample. Both 5′ gene expression libraries and V(D)J libraries were constructed following the manufacturer's standard 10× Genomics protocol (Single Cell 5 ′ Reagent Kits v5.2 User Guide) with modifications. Finally, paired-end 2×150 bp single-cell RNA and TCR V(D)J libraries were sequenced on a NovaSeq 6000 (Illumina).

scRNA-seq data processing and clustering

The scRNA-seq data were aligned and quantified using the CellRanger Single-Cell toolkit (v.6.1.1) against the GRCh38 human reference genome. To filter out low-quality cells, cells meeting the following criteria were used for downstream analysis: (i) expressing more than 200 genes; (ii) expressing fewer than 10,000 genes, and (iii) having less than 20% of the UMI mapped to the mitochondrial genome. Before analysis, DoubletFinder (v3; ref. 16) was used to remove doublet cells. A dissociation-associated gene (Supplementary Table S5) was applied to score cells, and cells with a score higher than 0.2 were filtered out to eliminate the effect of dissociation on cells. The detection rate of housekeeping genes was calculated, and cells with a detection rate below 40% were filtered out. Red blood cells were filtered out based on their gene expression, and cells with a proportion of more than 5% were filtered out. The top 2,000 highly variable genes identified from the “vst” method in the FindVariableFeatures function of Seurat were then subjected to principal component analysis (PCA). A shared nearest neighbor (SNN) graph was constructed using the top 10 principal components (PC), which were then used to cluster cells via the Louvain algorithm.

The resolution parameter was set to 0.8. The results were visualized using Uniform Manifold Approximation and Projection (UMAP), a kind of dimensionality reduction implemented by the RunUMAP function in Seurat. For each lineage cell type, the fastMNN pipeline (17) was used to remove batch effects between patients at the lineage level, and the Seurat pipeline was processed again using the default parameters except for the modulation of resolution and PCs to achieve sufficient fine clustering performance. The resolution was set to 0.8, 1.2, 0.4, 0.6, 0.8, and 0.4 for nonmalignant epithelial cells, ECs, PVL cells, T cells, B cells, and myeloid cells, respectively. The PCs were set to 15, 15, 15, 10, 10, and 10 for nonmalignant epithelial cells, ECs, PVL cells, T cells, myeloid cells, and B cells, respectively.

Cluster identity and annotation

First, we identified the main cell clusters with commonly used marker genes (MGP, PTN for PTN+ LP cells; TAGLN, ACAT2 for ACTA2+ Myo cells; KRT8 and KRT18 for KRT8+ ML cells; AGR2 and ERBB2 for malignant cells; KRT14 and KRT15 for KRT14+ basal cells/KCs; KRT1 and KRT10 for differentiating KCs; PMEL and DCT for melanocytes; CD3E and CD3D for T cells; NCAM1 and KLRD1 for NK cells; CD79A and MS4A1 for B cells; JCHAIN and MZB1 for plasma cells; LYZ and CD14 for myeloid cells; VWF and PECAM1 for ECs; RGS5 and MCAM for PVL cells; DCN and COL1A1 for fibroblasts). After that, we performed clustering analysis of each main cell cluster and annotated subclusters. Specifically, to confirm the identification of malignant cells, the copy number alteration of each cell was calculated. In this study, using fibroblasts, PVL cells, and ECs as the normal reference, inferCNV (https://github.com/broadinstitute/inferCNV) was applied to infer copy-number gain or loss events in epithelial cells. We scored all epithelial cells, fibroblasts, PVL cells, and ECs for the extent of CNV signal, defined as the mean of squares of CNV values across the genome. Putative malignant cells were defined as those with CNV signals above 0.08 and CNV correlations above 0.25. A more detailed bioinformatics analysis is provided in the supplementary data.

Multiplex immunofluorescence staining

FFPE blocks of IBC and non-IBC samples were cut into 4 μm sections. Staining was performed using a PANO 7-plex IHC Kit (Panovue, catalog no. 0004100100) according to the manufacturer's instructions. Briefly, the slides were incubated at 65°C for 2 hours for deparaffinization, hydration, and fixation. Antigen retrieval was carried out in EDTA (Solarbio, catalog no. C1034) or citrate (Solarbio, catalog no. C1032) buffer using a microwave (boiled then simmered for 15 minutes), followed by natural cooling. The slides were blocked with 10% goat serum (Bioss, catalog no. C-0005) at room temperature for 10 minutes. They were then incubated with different primary antibodies at 37°C for 30 minutes or overnight at 4°C in the proper order. After washing with TBST to remove excess antibodies, the slides were incubated with polymer HRP-anti-mouse/Rabbit IgG and subjected to tyramide signal amplification (TSA) for 10 minutes each at room temperature. This was followed by repeated antigen retrieval for the next round of staining or staining with 4′-6′-diamidino-2-phenylindole (DAPI; Panovue, catalog no. 0012100500) for 10 minutes. Finally, the slides were mounted for subsequent scanning and analysis. The antibodies used in each panel were as follows: Panel 1: Anti-EpCAM (Proteintech, catalog no. 66316–1-Ig, RRID: AB_2881697), anti-KRT8/18 (Abcam, catalog no. ab17139, RRID: AB_443679), anti-ITGA6 (Abcam, catalog no. ab181551, RRID: AB_2927695), and anti-PTN (Novus Biologicals, catalog no. H00005764-M01, RRID: AB_2174528). Panel 2: Anti-KRT8/18, anti-ITGA6, anti-PTN, anti-CD31 (Abcam, catalog no. ab28364, RRID: AB_726362), anti-CXCR4 (Abcam, catalog no. ab124824, RRID: AB_10975635), and anti-NRP1 (Abcam, catalog no. ab81321, RRID: AB_1640739). Panel 3: Anti-CD31, anti-CD146 (Abcam, catalog no. ab75769, RRID: AB_2143375), anti-NG2 (Abcam, catalog no. ab255811, RRID: AB_3076607), anti-αSMA (Abcam, catalog no. ab124964, RRID: AB_11129103), and anti-CD36 (Abcam, catalog no. ab133625, RRID: AB_2716564).

Multispectral imaging analysis

The TissueFAXS platform (TissueGnostics) was used to acquire multispectral images of the whole section at 20× magnification. The platform captures the fluorescence spectra at 20-nm wavelength intervals from 420 to 720 nm. Then, for each slide, a total of three to five fields of interest were taken based on their tissue sizes for qualitative digital analysis. Spectral unmixing was performed on multispectral images using StrataQuest software (version 7.1; TissueGnostics), which applied spectral libraries created from slides stained with individual reagents. This software also corrected autofluorescence by comparing the spectral profile of unstained slides to that of stained slides and subtracting the background accordingly for standardization across samples. Moreover, StrataQuest enabled the identification and quantification of cells with specific phenotypes using DAPI counterstaining for cell segmentation. This approach ensured standardized quantification for reliable comparative analysis across slides. The absolute percentage of cells demonstrating single marker or coexpression marker staining was assessed by a researcher who had no knowledge of the group assignment. The proportions of immature PVL cells (MCAM+CD36+) among total PVL cells, tip cells (CD31+CXCR4+) among ECs, and PTN+LP cells (KRT8/18+ITGA6+PTN+) among total epithelial cells (EPCAM+) or LP cells (KRT8/18+ITGA6+) were analyzed. Spatial colocalization analysis was meticulously performed on multispectral imaging data from reconstructed slides, with careful adjustments to preclude edge artifacts and data truncation. StrataQuest software was used to calculate the distance between cells of interest and the number of mutual neighbors (at the maximum distance of 160 μm) for each pair of cell phenotypes.

IHC

For IHC, 4-μm-thick slides cut from FFPE blocks were dewaxed and rehydrated by a decreasing ethanol series (100%, 90%, 80%, and 70% for 5 minutes each). The slides were blocked with 10% normal goat serum for 30 minutes, incubated with the following primary anti-PTN antibody (1:500; Novus Biologicals, catalog no. H00005764-M01, RRID: AB_2174528) overnight in a humidified chamber at 4°C, and subsequently probed with the secondary antibody (DAKO DAB Kit). The slides were scanned with a KF-PRO Slide Scanner (Kfbio), and the images were processed and analyzed with Image-Pro Plus (version 6.0, RRID:SCR_007369).

Enzyme-linked immunosorbent assay

Peripheral blood samples were collected in tubes without anticoagulant. The serum was obtained by centrifugation (3,500 rpm, 7 minutes), divided into aliquots, and immediately stored at −80°C until assayed. The culture supernatants of human umbilical vein ECs (HUVEC) stimulated with different concentrations of recombinant PTN protein were collected at 24 and 48 hours for detection. Serum PTN levels were quantified by ELISA with a commercial Human PTN ELISA Kit according to the manufacturer's instructions (Meimian, catalog no. MM-0264H1). Cell culture supernatant levels were quantified using a Human VEGFA ELISA Kit (Meimian, catalog no. MM-14698H1). Five different concentrations of the original density standard, ranging from 75 to 1,200 pg/mL, were prepared in standard diluent to establish the standard curve. The assay included a blank well and a positive well containing recombinant PTN protein (R&D Systems, catalog no. 252-PL-050) or recombinant VEGFA-165 protein (MedChemExpress, catalog no. HY-P78813). Serum samples (50 μL/well) were incubated in pre-coated 96-well microtiter plates for 0.5 hours at 37°C. After incubation, the plates were washed with washing buffer for 30 seconds, and this process was repeated five times. Complete removal of the liquid was ensured at each washing step. Next, 50 μL of the prepared horseradish peroxidase conjugate was added to each well at 37°C for 0.5 hours, except for the blank well. Afterward, the plates were washed five times with washing buffer, ensuring complete removal of the liquid at each step. Following the washes, 50 μL of chromogen solution A and 50 μL of chromogen solution B were added to each well, and the plates were kept in the dark at 37°C for 10 minutes. Finally, 50 μL of stop solution was added to each well. A microplate reader (Allsheng) was used to measure the absorbance at a wavelength of 450 nm, with the blank well serving as the zero reference. The standard curve was constructed by plotting known concentrations of the standards against their corresponding absorbance values. Linear regression analysis was then applied to derive the equation of the standard curve. The concentrations of unknown samples were determined by substituting their absorbance values into the standard curve equation.

Flow cytometry

Flow cytometry was performed on a CYTEK Aurora flow cytometer, and the results were analyzed using FlowJo v10.6.2 software (TreeStar). The following fluorochrome-conjugated antibodies were used for staining: anti-CD31-BUV496 (BD Biosciences, catalog no. 741146, RRID: AB_2870724) and anti-CXCR4-BV510 (BioLegend, catalog no. 306536, RRID: AB_2810460). For frozen single-cell resuscitation, the cells were incubated with 100 mg/mL DNase I at 37°C for 20 minutes to reduce cell adhesion, followed by resuspension in staining buffer (BD Bioscience, catalog no. 554657, RRID: AB_2869007). After that, the cells were incubated with fluorochrome-labeled antibodies at a dilution ratio of 1:100 in staining buffer for 20 minutes at room temperature. After the cells were washed and fixed, they were immediately analyzed.

Cell culture and transfection

HUVECs were purchased from Zhejiang Meisen Cell Technology (catalog no. CTCC-0804-PC). HUVECs were authenticated by the manufacturer and were tested free from Mycoplasma contamination by a PCR Mycoplasma Detection Kit (Abm). HUVECs were cultured in EC medium (ScienCell, catalog no. 1001) supplemented with 5% FBS, 1% EC growth supplement, and 1% penicillin–streptomycin. HUVECs were split every 3 to 4 days and used at passages 3 to 6. They were seeded in a 6‐well plate and cultivated for 24 hours, after which, the cells were transfected with NRP1 small-interfering RNA (siRNA) or a negative control (RiboBio) using Lipofectamine 3000 (Invitrogen). HUVECs were assayed 3 days after siRNA transfection. All cultures were maintained in a 5% CO2 cell culture incubator (Thermo Fisher Scientific) at 37°C.

HUVEC assays

For the migration assay, 1 × 105 HUVECs were seeded into the upper chamber of a transwell insert (pore size, 8 μm; Corning Falcon) in 400 μL of serum-free DMEM (Gibco). The lower chambers were supplemented with 600 μL of 5% FBS supplemented with different concentrations (0, 10, 100, 300 ng/mL) of recombinant PTN protein (R&D Systems, catalog no. 252-PL-050). For siRNA treatment, 100 ng/mL PTN was added to the lower chamber. To examine whether the function of PTN in angiogenesis depends on VEGFA and VEGF-related pathways, HUVECs were treated with bevacizumab (200 ng/mL; MedChemExpress, catalog no. HY-P9906) or vandetanib (5 μm; Selleck Chemicals, catalog no. S1046) in the upper chamber. To investigate whether VEGFA competes with PTN for the NRP1 receptor, the concentration of recombinant VEGFA protein (MedChemExpress, catalog no. HY-P78813) in the lower chamber was 100 ng/mL. In the upper chamber, HUVECs or HUVECs treated with siRNA were added and suspended in DMEM with or without vandetanib (5 μm). After 12 hours, the chambers were collected and stained with 0.5% crystal violet, followed by observation using a light microscope (Olympus) and counting in five random fields.

For the tube formation assay, 96-well plates were precoated with 60 μL of Matrigel containing different concentrations (0, 10, 100, or 300 ng/mL) of recombinant PTN protein at 37°C for 1 hour. For siRNA treatment, the concentration of PTN was 100 ng/mL. HUVECs (2 × 104) were resuspended in 100 μL of conditioned medium to be detected, added to a Matrigel-coated well and grown in a 5% CO2 cell culture incubator (Thermo Fisher Scientific) at 37°C for 4 to 6 hours. To determine whether the function of PTN in angiogenesis depends on VEGFA and VEGF-related pathways, HUVECs were treated with bevacizumab (200 ng/mL) or vandetanib (5 μm) in the upper chamber. To investigate whether VEGFA competes with PTN for the NRP1 receptor, the concentration of VEGFA in the Matrigel-coated wells was 100 ng/mL. HUVECs or HUVECs treated with siRNAs were suspended in conditioned medium with or without vandetanib (5 μm). Analysis was performed with an IX83 inverted microscope (Olympus) and ImageJ.

HUVEC sprouting assays were performed as described previously (18). The concentration gradient of the recombinant PTN protein in the culture medium was as described previously.

RT-qPCR

Total RNA was extracted from the cells using TRizol reagent (Invitrogen). After RNA reverse transcription via a Reverse Transcription Master Kit (Takara) using 2 μg of total RNA in a 20 μL reaction system, qPCR analysis was performed with SYBR Green (SYBR Premix Ex Taq II; Takara) in a QuantStudio 7 Flex Real-Time PCR System (Applied Biosystems). The analysis of each sample was repeated three times, and the gene expression levels were calculated using the 2−ΔΔCT method, with GAPDH as the normalization control.

Western blot analysis

The first step involved preparing cell lysates from the NRP1 knockdown and negative control groups using RIPA lysis buffer (EpiZyme, catalog no. PC101). Protease inhibitor (Roche, catalog no. 5892970001) and phosphatase inhibitors (Roche, catalog no. 04906845001) were added to the lysis buffer, and all manipulations were performed on ice. The lysate was stored at −20°C or for long-term storage at −80°C. The protein concentration was determined using a detergent-compatible protein assay (EpiZyme, catalog no. ZJ101). Subsequently, 20 μg of protein was separated on a gel comprising a 10% upper layer and a 12.5% lower layer and transferred onto polyvinylidene fluoride membranes (Millipore, catalog no. IPVH00010). The membranes were then incubated with 5% milk at room temperature for 1 hour. Next, the membranes were incubated with the appropriate primary antibodies overnight at 4°C. After washing three times in TBST (EpiZyme, catalog no. PS103S), the blots were incubated with the HRP-conjugated secondary antibody anti-rabbit IgG (1:10,000; Cell Signaling Technology, catalog no. 7074, RRID: AB_2099233) at room temperature for 1 hour. Proteins of interest were visualized using an Omni-ECL Efficient Light Chemiluminescence Kit (EpiZyme, catalog no. SQ203), with β-actin serving as the total protein loading control. The membrane was incubated at 4°C overnight with the following antibodies: anti-NRP1 (1:1,000, Abcam, catalog no. ab81321, RRID: AB_1640739) and anti-β-actin (1:1,000, Cell Signaling Technology, catalog no. 4970S, RRID: AB_2223172).

Survival analysis

Bulk RNA-seq data were used to evaluate the prognostic significance of tip cell levels in the TCGA, Molecular Taxonomy of Breast Cancer International Consortium (METABRIC), and Population-Based Multicenter Sweden Cancerome Analysis Network – Breast (SCAN-B) Initiative cohorts. The infiltration level of tip cells was calculated using the gsva method from the gene set variation analysis (GSVA) R package (version 1.32.0) based on the marker genes identified from the scRNA-seq data (Supplementary Table S6). Patients were grouped into tip cell high expression and tip cell low-expression groups according to the GSVA score of the tip cell markers, and the optimal cutoff value was calculated by the surv_cutpoint function of the survminer R package (version 0.5.9). Subsequently, Kaplan–Meier survival curves were plotted using the R function ggsurvplot to compare overall survival between patients in the high tip cell expression group and those in the low tip cell expression group. In addition, the correlation between PTN expression and overall survival was also evaluated in the TCGA cohort. Patients were grouped into PTN-high expression and PTN-low-expression groups according to the transcripts per kilobase million expression of PTN. The optimal cutoff value was also calculated by the surv_cutpoint function of the survminer R package (version 0.5.9), and Kaplan–Meier survival curves were also plotted using the R function ggsurvplot.

Study approval

This study was centrally approved by the Ethical Committee of the First Affiliated Hospital of Sun Yat-sen University (No. [2022]135) and complied with all relevant ethical regulations. All patients provided written informed consent for the samples used for research.

Data availability

The raw-sequencing data generated in this study, including the scRNA-seq, bulk WES, and RNA-seq data, are publicly available in the Genome Sequence Archive (GSA) for Humans (https://ngdc.cncb.ac.cn/gsa-human/) under accession number HRA006297. The raw data download can be requested according to GSA's instructions, and all requests for data will be approved after review by its Data Access Committee. The publicly available scRNA-seq data analyzed in this study were obtained from the Gene Expression Omnibus (GEO) at GSE176078 (11) and GSE161529 (19). The processed scRNA data of EGAD00001006608 in the European Genome-phenome Archive (EGA) were downloaded from https://lambrechtslab.sites.vib.be/en/single-cell (20). The RNA microarray data of IBC were downloaded from GEO at GSE22597. The bulk RNA-seq data and clinical information of the TCGA, METABRIC and SCAN-B cohorts were obtained from the Genomic Data Commons Data Portal (https://portal.gdc.cancer.gov/), cBioPortal (https://www.cbioportal.org/study/summary?id=brca_metabric), and GEO at GSE96058, respectively. The raw WES data generated by the TCGA for non-IBC patients were downloaded from the Genomic Data Commons Data Portal. All other raw data generated in this study are available upon request from the corresponding author.

Statistical analysis

Statistical analyses were performed using R (version 4.1.2, RRID: SCR_001905). Statistical significance was defined as a P value less than 0.05 according to two-sided Wilcoxon tests and t tests, as described in the figure legends. For the experimental data, GraphPad Prism 9 (GraphPad Software, RRID: SCR_002798) was used to perform the statistical analyses and graphics production.

Single-cell transcriptional profiling defines the cellular ecosystem of IBC

To investigate the cellular diversity and molecular signatures in IBC, we performed scRNA-seq on tissues derived from paired tumors and skin with erythema and edema (referred to as “affected skin”) from 17 treatment-naïve patients with IBC, as well as from tumors, skin with tumors, and normal skin from 5 treatment-naïve patients with non-IBC (Fig. 1A; Supplementary Table S1). Moreover, three independent published scRNA-seq datasets of non-IBC tumor and normal breast tissue samples were included for integrated analysis. Another independent cohort of 52 patients with IBC was used to validate our transcriptional findings (Fig. 1A; Supplementary Table S2).

Figure 1.

Single-cell transcriptional profiling defines the cellular ecosystem of IBCs and non-IBCs. A, Workflow of the study design. Left, discovery cohort and the integration analyses included public data from the non-IBC_Tumor-KULEUVEN, non-IBC_Tumor-TKCC, non-IBC_Tumor-WEHI, and Normal_Breast-WEHI cohorts and the TCGA database. Right, validation cohorts and in vitro experiments. The numbers (n) of patients and samples are provided in the figure. KULEUVEN, KU Leuven; TKCC, the Kinghorn Cancer Centre; WEHI, the Walter and Eliza Hall Institute; MDA, MD Anderson Cancer Center. B, The UMAP plots of immune cells, stromal cells, and epithelial cell clusters from IBC tumors (IBC_Tumor), affected skin of IBC (IBC_Skin), non-IBC tumors (non-IBC_Tumor), non-IBC skin with tumors (non-IBC_Skin) and normal skin (Normal_Skin; left), and UMAPs projected by tissue origin and group (right). C, Box plots showing the CNV scores of malignant cells, nonmalignant epithelial cells, and stromal cells (ECs, PVL cells, and fibroblasts) in both tumor and skin samples. Wilcoxon rank-sum test; ***, P < 0.001. D, Distributional preference of each cell cluster among each group, calculated by the Ro/e score. Chi-square test; ***, P < 0.001; ****, P < 0.0001. E, Bar chart showing the proportions of epithelial cell clusters in 16 IBC tumors and 5 non-IBC tumors. F, Dot plot showing the expression of the top five differentially expressed genes in each epithelial cluster.

Figure 1.

Single-cell transcriptional profiling defines the cellular ecosystem of IBCs and non-IBCs. A, Workflow of the study design. Left, discovery cohort and the integration analyses included public data from the non-IBC_Tumor-KULEUVEN, non-IBC_Tumor-TKCC, non-IBC_Tumor-WEHI, and Normal_Breast-WEHI cohorts and the TCGA database. Right, validation cohorts and in vitro experiments. The numbers (n) of patients and samples are provided in the figure. KULEUVEN, KU Leuven; TKCC, the Kinghorn Cancer Centre; WEHI, the Walter and Eliza Hall Institute; MDA, MD Anderson Cancer Center. B, The UMAP plots of immune cells, stromal cells, and epithelial cell clusters from IBC tumors (IBC_Tumor), affected skin of IBC (IBC_Skin), non-IBC tumors (non-IBC_Tumor), non-IBC skin with tumors (non-IBC_Skin) and normal skin (Normal_Skin; left), and UMAPs projected by tissue origin and group (right). C, Box plots showing the CNV scores of malignant cells, nonmalignant epithelial cells, and stromal cells (ECs, PVL cells, and fibroblasts) in both tumor and skin samples. Wilcoxon rank-sum test; ***, P < 0.001. D, Distributional preference of each cell cluster among each group, calculated by the Ro/e score. Chi-square test; ***, P < 0.001; ****, P < 0.0001. E, Bar chart showing the proportions of epithelial cell clusters in 16 IBC tumors and 5 non-IBC tumors. F, Dot plot showing the expression of the top five differentially expressed genes in each epithelial cluster.

Close modal

For scRNA-seq, a total of 142,841 epithelial cells, 66,598 immune cells and 36,436 stromal cells were successfully harvested for dimensionality reduction and UMAP analysis (Fig. 1B; Supplementary Fig. S1A; Supplementary Table S7). Moreover, copy-number variation (CNV) analysis was also employed to distinguish between malignant and nonmalignant epithelial cells with a CNV signal threshold above 0.08 and a CNV correlation above 0.25 (Fig. 1C; Supplementary Figs. S1B–S1D; Supplementary Table S7; ref. 21). As a result, unbiased clustering identified 16 distinct cell types (11, 20, 22), including malignant cells, PTN+ LP cells, ACTA2+ myoepithelial (Myo) cells, KRT8+ mature luminal (ML) cells, KRT14+ basal cells, KRT1+ differentiating keratinocytes (KC), KRT14+ basal KCs, B cells, plasma cells, T cells, myeloid cells, natural killer (NK) cells, fibroblasts, PVL cells, ECs, and melanocytes, based on their marker gene expression (Fig. 1B; Supplementary Fig. S1E; Supplementary Table S8). Cells in each cluster were distributed in different samples without patient-specific clusters (Supplementary Fig. S1F).

To further analyze the microenvironment of IBC, the enrichment score was quantified by calculating the ratio of observed to expected cell numbers in each cluster (Ro/e). Interestingly, we found that a nonmalignant epithelial cell subpopulation, PTN+ LP cells, was the most significantly enriched subpopulation in IBC tumors (Fig. 1D), comprising 2.84% to 71.26% of epithelial cells across IBC tumors and only 0.04% to 0.96% of those across non-IBC tumors (Fig. 1E; Supplementary Fig. S1G). The PTN+ LP subpopulation was characterized by high expression of the PTN gene (Fig. 1F; Supplementary Table S9), which encodes an 18-kDa secreted protein and acts as an important developmental factor for central nervous system embryogenesis (23). However, the expression of PTN was extremely low in malignant cells (Fig. 1F).

Because of the diffuse erythema phenotype and edema of the overlying skin, we initially explored the immune cell composition across tumor and skin samples from all patients with IBC and patients with non-IBC on the basis of established gene signatures (11, 24, 25). First, we identified four clusters of CD8+ T cells, four clusters of CD4+ T cells and two clusters of NK cells (Supplementary Figs. S1H and S1I). Among the four CD8+ clusters, one cluster (CD8_C8_PDCD1) exhibited high expression of inhibitory checkpoint molecules, including LAG3, TIGIT, PDCD1, and HAVCR2, and another cluster (CD8_C2_GZMK) exhibited lower PDCD1 expression with relatively high levels of cytotoxic markers, including GZMK, GZMA, IFNG, and NKG7 (Supplementary Fig. S1I). The other two clusters exhibited a cytotoxic signature (CD8_C6_CX3CR1) or a naïve signature (CD8_C0_ANXA1; Supplementary Fig. S1I). CD4 clusters consisted of Treg cells marked by IL2RA and IKZF2 (CD4_C3_FOXP3), naïve/central memory CD4+ T cells (CD4_C5_CXCR7, CD4_C1_GPR183), and exhausted CD4+ T cells (CD4_C4_CXCL13). We also identified two NK-cell clusters (NK_C7_FCGR3A and NK_C9_AREG) by the expression of NK-cell markers (NCAM1, TYROBP, and NKG7) (Supplementary Fig. S1I). Among these clusters, CD8_C8_PDCD1, CD4_C1_GPR183, CD4_C4_CXCL13, and CD4_C5_CCR7 cells were enriched in IBC tumors, whereas CD8_C6_CX3CR1, NK_C7_FCGR3A, and NK_C9_AREG cells were enriched in the affected skin of patients with IBC compared with those in the skin of patients with non-IBC (Supplementary Fig. S1H).

Next, myeloid cells were defined as containing two clusters of monocytes, four clusters of macrophages, and five clusters of dendritic cells (DC; Supplementary Fig. S1J and S1K). Among them, a higher proportion of Mac_C0_C1QB and Mac_C2_MARCO were observed in IBC tumors than in non-IBC tumors (Supplementary Fig. S1J). Interestingly, the affected skin of patients with IBC showed increased levels of all macrophage clusters (Supplementary Fig. S1J). In addition, two clusters of B cells and one cluster of plasma cells were identified in the B-cell compartment (Supplementary Fig. S1L). We speculated that the abundant infiltration and alteration of immune cells result from the high vascularization of IBCs. Therefore, investigating the specific alterations in the breast parenchyma may be the key to unraveling the unique presentation of IBC. Collectively, these data provide insights into the heterogeneity of the immune microenvironment in tumors and the affected skin of patients with IBC and may guide immunotherapeutic strategies for this disease.

Enrichment of the PTN+ LP subpopulation in IBC tumors

On the basis of the above results, we further examined PTN+ LP cells in non-IBC tissues by integrating three published single-cell transcriptomic datasets (Supplementary Table S3; refs. 11, 19, 20) with our scRNA-seq dataset. CNV analysis was also performed to distinguish malignant from nonmalignant epithelial cells (Supplementary Fig. S2A; Supplementary Table S10). We obtained a total of 19,434 nonmalignant epithelial cells and integrated them with our scRNA-seq data (Supplementary Fig. S2B), and five major subpopulations were observed in IBC and non-IBC tumors. These clusters were identified on the basis of known representative genes (19, 22) and their marker genes, including (i) KRT14+ basal cells (KRT14, KRT5), (ii) PTN+ LP cells (PTN, SLPI, ITGA6, KIT, PROM1), (iii) PTN LP cells (SLPI, ITGA6, KIT, PROM1), (iv) KRT8+ ML cells (MUC1, MGP), and (v) ACTA2+ Myo cells (ACTA2, TAGLN; Fig. 2A; Supplementary Figs. S2B and S2C). Consistently, the integrated data also revealed that the PTN+ LP subpopulation was more abundant in IBC tumors, whereas PTN LP cells were enriched in non-IBC tumors (Fig. 2B; Supplementary Fig. S2C). Moreover, we analyzed RNA microarray data of 25 patients with IBC (referred to as “IBC_Tumor-MDA”) from GEO databases (GSE22597) and found that the gene signature score of PTN+ LP cells in IBC tumors was significantly higher than that in non-IBC tumors (Fig. 2C). In addition, the expression level of PTN was significantly higher in IBC-LP cells than in non-IBC-LP cells (Fig. 2D). Furthermore, we performed bulk RNA-seq and observed that the expression of PTN in IBC tumors was higher than that in non-IBC tumors (Fig. 2E; Supplementary Table S1).

Figure 2.

Enrichment of the PTN+ LP subpopulation in IBC tumors. A, UMAP plot of the combination of nonmalignant epithelial cell clusters from our cohort and those from three independent published cohorts (non-IBC_Tumor-KULEUVEN, non-IBC_Tumor-TKCC, and non-IBC_Tumor-WEHI). B, Distributional preference of each nonmalignant epithelial cluster among IBC_Tumor, non-IBC_Tumor, non-IBC_Tumor-KULEUVEN, non-IBC_Tumor-TKCC, and non-IBC_Tumor-WEHI. Distributional preference of each cluster among each group, as calculated by the Ro/e score and Chi-square test; ****, P < 0.0001. C, Box plots showing the difference in the scores of PTN+ LP cells between IBC_Tumor and non-IBC tumors from one independent published cohort (IBC_Tumor-MDA). Wilcoxon rank-sum test; ***, P < 0.05. D, Comparison of PTN expression in LP cells between the IBC tumor and non-IBC tumor from all cohorts. Wilcoxon rank-sum test; ***, P < 0.001. E, Volcano plot showing the differentially expressed genes between IBC tumors (red dots) and non-IBC tumors (blue dots) from our bulk RNA-seq paired with our scRNA-seq samples. F, UMAP plot of the combination of nonmalignant epithelial cells in IBC tumors from our dataset and epithelial cells in normal breast tissue from one independent dataset (Normal_breast-WEHI). G, Distributional preference of each epithelial cell cluster between IBC tumor and normal breast tissue. ****, P < 0.0001. H,PTN expression in LP cells between IBC tumor tissue and normal breast tissue. Wilcoxon rank-sum test. ***, P < 0.001. I, Representative images of mIF staining for PTN+ LP cells (KRT8/18+ ITGA6+ PTN+, red arrowheads) in IBC-adjacent normal tissues (IBC_Adjacent). Scale bars, 20 μm. J, The percentage (left) and density (right) of PTN+ LP cells in IBC and non-IBC adjacent normal tissues. The number of dots indicates the number of patients. Wilcoxon rank-sum test; ****, P < 0.0001. K, Representative images of IHC staining for PTN in adjacent normal tissues and IBC tumors (left) and non-IBC tissues (right). Scale bars, 20 μm. L, Quantitative integrated optical density (IOD) analysis of the level of PTN by IHC. The number of dots indicates the number of patients. The number of dots indicates the number of patients. Unpaired t test; ***, P < 0.001. M, PTN serum levels of patients with IBC and non-IBC. The number of dots indicates the number of patients. Wilcoxon rank-sum test; *, P < 0.05. N, Bar plot showing the enriched Gene Ontology pathways in PTN+ LP cells compared with PTN LP cells.

Figure 2.

Enrichment of the PTN+ LP subpopulation in IBC tumors. A, UMAP plot of the combination of nonmalignant epithelial cell clusters from our cohort and those from three independent published cohorts (non-IBC_Tumor-KULEUVEN, non-IBC_Tumor-TKCC, and non-IBC_Tumor-WEHI). B, Distributional preference of each nonmalignant epithelial cluster among IBC_Tumor, non-IBC_Tumor, non-IBC_Tumor-KULEUVEN, non-IBC_Tumor-TKCC, and non-IBC_Tumor-WEHI. Distributional preference of each cluster among each group, as calculated by the Ro/e score and Chi-square test; ****, P < 0.0001. C, Box plots showing the difference in the scores of PTN+ LP cells between IBC_Tumor and non-IBC tumors from one independent published cohort (IBC_Tumor-MDA). Wilcoxon rank-sum test; ***, P < 0.05. D, Comparison of PTN expression in LP cells between the IBC tumor and non-IBC tumor from all cohorts. Wilcoxon rank-sum test; ***, P < 0.001. E, Volcano plot showing the differentially expressed genes between IBC tumors (red dots) and non-IBC tumors (blue dots) from our bulk RNA-seq paired with our scRNA-seq samples. F, UMAP plot of the combination of nonmalignant epithelial cells in IBC tumors from our dataset and epithelial cells in normal breast tissue from one independent dataset (Normal_breast-WEHI). G, Distributional preference of each epithelial cell cluster between IBC tumor and normal breast tissue. ****, P < 0.0001. H,PTN expression in LP cells between IBC tumor tissue and normal breast tissue. Wilcoxon rank-sum test. ***, P < 0.001. I, Representative images of mIF staining for PTN+ LP cells (KRT8/18+ ITGA6+ PTN+, red arrowheads) in IBC-adjacent normal tissues (IBC_Adjacent). Scale bars, 20 μm. J, The percentage (left) and density (right) of PTN+ LP cells in IBC and non-IBC adjacent normal tissues. The number of dots indicates the number of patients. Wilcoxon rank-sum test; ****, P < 0.0001. K, Representative images of IHC staining for PTN in adjacent normal tissues and IBC tumors (left) and non-IBC tissues (right). Scale bars, 20 μm. L, Quantitative integrated optical density (IOD) analysis of the level of PTN by IHC. The number of dots indicates the number of patients. The number of dots indicates the number of patients. Unpaired t test; ***, P < 0.001. M, PTN serum levels of patients with IBC and non-IBC. The number of dots indicates the number of patients. Wilcoxon rank-sum test; *, P < 0.05. N, Bar plot showing the enriched Gene Ontology pathways in PTN+ LP cells compared with PTN LP cells.

Close modal

To explore whether PTN+ LP cells also exist in normal mammary glands, we analyzed published scRNA-seq data from 11 healthy women (Supplementary Table S4; ref. 19) and observed the major epithelial cell types in normal breast glands, including PTN LP cells, KRT8+ ML cells, KRT14+ basal cells, and ACTA2+ Myo cells (Fig. 2F; Supplementary Figs. S2D and S2E). As expected, the PTN+ LP subpopulation was significantly enriched in IBCs with a higher expression level of PTN compared with that in normal mammary glands (Fig. 2G and H; Supplementary Fig. S2E).

Next, multiplex immunofluorescent (mIF) staining of tumor-adjacent regions from patients with IBC confirmed the existence and enrichment of PTN+ LP cells in IBCs via the coexpression of KRT8/18, ITGA6, and PTN (Fig. 2I and J; Supplementary Fig. S2F). IHC was performed on tumor tissues and adjacent normal tissues to validate the expression of PTN in IBC. Notably, the expression of PTN in adjacent normal regions in patients with IBC was much higher than that in patients with non-IBC, whereas the expression of PTN in tumor regions was lower in both patients with IBC and patients with non-IBC (Fig. 2K and L). Moreover, we performed an ELISA to measure the serum levels of PTN in 26 patients with IBC and 30 patients with non-IBC. We found that the serum concentrations of PTN were significantly higher in patients with IBC than in patients with non-IBC (Fig. 2M), suggesting the presence of large amounts of secretory PTN in patients with IBC. Furthermore, among patients with IBC, PTN expression levels in adjacent normal tissues were positively correlated with PTN concentrations in serum (Supplementary Fig. S2G), suggesting that PTN may serve as a potential circulating biomarker. Pathway enrichment analysis revealed that EC migration, cell migration involved in sprouting angiogenesis, and angiogenesis were upregulated in IBC tumors (Fig. 2N; Supplementary Table S11). Therefore, we sought to explore whether PTN+LP cells could regulate angiogenesis and contribute to the unique IBC-specific microenvironment.

PTN+ LP interacts with ECs via the PTN–NRP1 axis

The finding that PTN+ LP cells may play a proangiogenic role in IBC prompted us to further focus on ECs. To this end, we first performed unsupervised subclustering of ECs, and six subpopulations were identified by using our scRNA-seq dataset. On the basis of the expression of known representative genes (26, 27), we identified arterial (GJA5), capillary (CA4), venous (ACKR1), lymphatic (PROX1), tip (CXCR4), and immature (APLNR) ECs in IBCs and non-IBCs (Fig. 3A; Supplementary Figs. S3A and S3B; Supplementary Table S6). Notably, tip cells, which are referred to as the traditional angiogenic phenotype (26, 28), were enriched in both IBC tumors and skin compared with those in non-IBC tumors, non-IBC skin with tumors, and normal skin (Fig. 3B). Next, we assessed whether PTN is associated with angiogenesis in IBC by using single-sample GSEA to calculate the score of tip cells based on bulk RNA-seq data paired with our scRNA-seq data. The results showed that the normalized PTN gene expression in the tumor and affected skin of patients with IBC was significantly positively correlated with the score of tip cells (Fig. 3C), suggesting that PTN plays a critical role in angiogenesis. This result was consistent with that observed in the IBC-Tumor_MDA cohort (GSE22597; Supplementary Fig. S3C). These findings prompted us to investigate the potential interaction between PTN+ LP cells and ECs.

Figure 3.

PTN+ LP cells interact with ECs via the PTN–NRP1 axis. A, UMAP plot of identified EC clusters from our scRNA-seq data. B, Heatmap displaying the relative enrichment of each EC cluster between IBC_Tumor and non-IBC_Tumor (left) and among IBC_Skin, non-IBC_Skin, and Normal-Skin (right), as calculated by the Ro/e score. Chi-square test; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. C, Correlation between PTN expression and the score of tip cell in IBC_Tumor (left) and IBC_Skin (right) samples according to bulk RNA-seq data. Spearman rank correlation coefficient was calculated to infer statistical significance. D, Dot plots showing the ligand–receptor interactions of the PTN pathway between PTN+ LP cells and EC subtypes in tumors and skin. The dot size represents the P value calculated by the permutation test, colored by interaction strength. E, Representative mIF staining images showing adjacent normal IBC tissues. White arrowheads, interaction between PTN secreted by PTN+ LP cells (KRT8/18+ ITGA6+ PTN+, labeled with red arrowheads) and NRP1 on tip cells (CD31+ CXCR4+ NRP1+, labeled with yellow arrowheads). Scale bars, 20 μm. F, Representative mIF staining images of the affected skin of IBC patients. The white arrowheads indicate the interaction between PTNs secreted by PTN+ LP cells and NRP1 on tip cells (CD31+CXCR4+NRP1+, yellow arrowheads). Scale bars, 20 μm. G, Percentages and densities of tip cells (CD31+ CXCR4+) in the IBC_Adjacent and non-IBC_Adjacent groups and in the IBC_Skin and non-IBC_Skin groups. Wilcoxon rank-sum test; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. H, Comparison of the distance (distance <160 μm) between PTN+ LP cells and NRP1+ tip cells and the distance between PTNLP cells and NRP1+ tip cells. Wilcoxon rank-sum test. ****, P < 0.0001.

Figure 3.

PTN+ LP cells interact with ECs via the PTN–NRP1 axis. A, UMAP plot of identified EC clusters from our scRNA-seq data. B, Heatmap displaying the relative enrichment of each EC cluster between IBC_Tumor and non-IBC_Tumor (left) and among IBC_Skin, non-IBC_Skin, and Normal-Skin (right), as calculated by the Ro/e score. Chi-square test; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. C, Correlation between PTN expression and the score of tip cell in IBC_Tumor (left) and IBC_Skin (right) samples according to bulk RNA-seq data. Spearman rank correlation coefficient was calculated to infer statistical significance. D, Dot plots showing the ligand–receptor interactions of the PTN pathway between PTN+ LP cells and EC subtypes in tumors and skin. The dot size represents the P value calculated by the permutation test, colored by interaction strength. E, Representative mIF staining images showing adjacent normal IBC tissues. White arrowheads, interaction between PTN secreted by PTN+ LP cells (KRT8/18+ ITGA6+ PTN+, labeled with red arrowheads) and NRP1 on tip cells (CD31+ CXCR4+ NRP1+, labeled with yellow arrowheads). Scale bars, 20 μm. F, Representative mIF staining images of the affected skin of IBC patients. The white arrowheads indicate the interaction between PTNs secreted by PTN+ LP cells and NRP1 on tip cells (CD31+CXCR4+NRP1+, yellow arrowheads). Scale bars, 20 μm. G, Percentages and densities of tip cells (CD31+ CXCR4+) in the IBC_Adjacent and non-IBC_Adjacent groups and in the IBC_Skin and non-IBC_Skin groups. Wilcoxon rank-sum test; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. H, Comparison of the distance (distance <160 μm) between PTN+ LP cells and NRP1+ tip cells and the distance between PTNLP cells and NRP1+ tip cells. Wilcoxon rank-sum test. ****, P < 0.0001.

Close modal

Because of the insufficient study of cell surface receptors for PTN (29), we investigated the known receptors for PTN, including PTPRZ1, NCL, SDC3, ITGAV, ITGB3, and NRP1. Ligand–receptor (L–R) algorithms revealed strong cross-talk between PNT+ LP cells and ECs in tumors and the affected skin of patients with IBC via PTN and its receptor NRP1 (Fig. 3D; Supplementary Fig. S3D). Previous studies have also demonstrated that NRP1 is a receptor for PTN that modulates cell trafficking and signaling pathways and thus enhances cell invasion, angiogenesis, and tumor progression (30). To validate this interaction, we performed mIF staining on tumor and skin samples and found that tip cells (CD31+CXCR4+) were enriched in the tumors and the affected skin of patients with IBC (Fig. 3EG; Supplementary Figs. S3E and S3F). In addition, flow cytometry analyses demonstrated an increased proportion of tip cells in both the tumors and affected skin of patients with IBC (Supplementary Figs. S3G and S3H). The spatial analysis of LP cells and NRP1+ tip cells showed that PTN+ LP cells (KRT8/18+ ITGA6+ PTN+) were significantly closer to NRP1+ tip cells (CD31+ CXCR4+ NRP1+) than PTN LP cells (KRT8/18+ITGA6+) in the adjacent normal tissues of patients with IBC (Fig. 3H). We further utilized mIF staining to quantify the proportions of PTN+ LP cells and NRP1+ tip cells and detected a positive correlation between the proportions of these two types of cells (Supplementary Fig. S3I). To evaluate the association of tip cells with the outcome of patients with breast cancer, we used the bulk RNA-seq datasets from the TCGA, METABRIC, and SCAN-B to evaluate the enrichment of the gene set signatures of tip cells via GSVA. The results showed that patients with breast cancer expressing high-level tip cell gene signatures had shorter overall survival, suggesting that patients with breast cancer with active angiogenesis had a worse prognosis (Supplementary Fig. S3J).

VEGF is a well-established cytokine that plays a key role in the development of tumor blood vessels, including those in breast cancer (31). However, the efficacy of anti-VEGFA therapy is limited in patients with IBC (32). We thus examined whether PTN secreted by PTN+ LP cells promotes the formation of massive blood vessels via VEGF in IBC tumors. Interestingly, our correlation analysis of our bulk RNA-seq data and public data (IBC_Tumor-MDA, GSE22597) revealed no correlation between the expression of PTN and the VEGFA/VEGF signaling pathway (Supplementary Figs. S3K and S3L; Supplementary Table S12). We further extracted ECs from the scRNA-seq data and used weighted gene coexpression network analysis (WGCNA) to identify genes coexpressed with NRP1. Interestingly, there was no coexpression of NRP1 and VEGF-related genes according to Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis (Supplementary Fig. S3M; Supplementary Table S12).

Overall, our findings suggest that PTN secreted by LP cells could boost angiogenesis in a VEGFA-independent manner in both the tumors and affected skin of patients with IBC.

Subtypes of PVL cells and their functional characteristics

During angiogenesis, PVL cells are recruited by tip cells to support vascular stability through close contact and signaling cross-talk with the endothelium and play a contractile role in regulating vascular stability (33, 34). In addition, previous studies have shown that PVL cells are associated with breast cancer metastasis (12, 35). Hence, we delineated the phenotypes of PVLs to further explore the contributions of cell communication to angiogenic processes and possible functions in IBC. As a result, four clusters of PVL cells, namely, PVL_MYH11, PVL_MMP9, PVL_CCL19, and PVL_IGFBP3, were defined (Fig. 4A; Supplementary Figs. S4A–S4C; Supplementary Table S13). Among these clusters, the PVL_MYH11 cluster was characterized by the enrichment of myogenic differentiation genes (MYH11, MYLK, TAGLN, and ACTA2; Supplementary Fig. S4C). The PVL_MMP9, PVL_CCL19, and PVL_IGFBP3 clusters were characterized by elevated expression of genes associated with immature and stem phenotypes (PDGFRB, RGS5, CD36, and CD44; Supplementary Fig. S4C). On the basis of these observations, we constructed a developmental trajectory using Monocle 2 and inferred dynamic differentiation-state transition processes from PVL_IGFBP3, PVL_MMP9, and PVL_CCL19 to PVL_MYH11 (Fig. 4B). The expression of MYH11, TAGLN, MYLK, and ACTA2 gradually increased over pseudotime and peaked at the end of pseudotime period, whereas decreasing trends were observed for canonical markers related to stem cells (CD44), immature phenotypes (RGS5, CD36, PDGFRB, and CD248) and adhesion molecules (ICAM1, ITGA1, and VCAM1; refs. 11, 12) along the pseudotime trajectory (Supplementary Fig. S4D).

Figure 4.

Subtypes of PVL cells and their functional characteristics. A, UMAP plot of identified PVL cell clusters from our scRNA-seq data. B, Pseudotime trajectories of different PVL cell clusters calculated by Monocle 2. Different PVL cell clusters are labeled with different colors. C, Interaction networks showing the cross-talk weights between PVL cell and EC clusters in IBC_Tumor and IBC_Skin. The band width is proportional to the interaction number. D, Dot plots showing PDGF pathway ligand–receptor interactions between tip cells and PVL cell clusters in each group. The dot size represents the P value calculated by the permutation test, colored by interaction strength. E, Dot plots showing ANGPT pathway ligand–receptor interactions between tip cells and EC clusters in each group. The dot size represents the P value calculated by the permutation test, colored by interaction strength. F, Heatmaps displaying the relative enrichment of each PVL cell cluster among IBC_Skin, non-IBC_Skin, and Normal_Skin (top) and between IBC_Skin and IBC_Tumor (bottom), as calculated by the Ro/e score. Chi-square test; ****, P < 0.0001. G, Representative mIF images for validation of the location of ECs (CD31+), immature PVL cells (MCAM+ CD36+, white arrowheads), and differentiated PVL cells (αSMA+ NG2+ MCAM+, red arrowheads) in the affected skin of patients with IBC. Scale bars, 20 μm. H, Representative mIF images for validation of the location of ECs (CD31+) and differentiated PVL cells (αSMA+ NG2+ MCAM+, red arrowheads) in IBC_Tumors. Scale bars, 20 μm. I, Top, the percentages of immature PVL cells (CD146+ CD36+) in the PVL cell population were compared among the IBC_Skin, non-IBC_Skin, and IBC_Tumor groups. Bottom, the densities of immature PVL cells were compared among IBC_Skin, non-IBC_Skin, and IBC_Tumor. Wilcoxon rank-sum test; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. J, Bar plot showing the enriched GO pathways in PVL_MMP9 and PVL_CCL19 compared with PVL_IGFBP3 and PVL_MYH11.

Figure 4.

Subtypes of PVL cells and their functional characteristics. A, UMAP plot of identified PVL cell clusters from our scRNA-seq data. B, Pseudotime trajectories of different PVL cell clusters calculated by Monocle 2. Different PVL cell clusters are labeled with different colors. C, Interaction networks showing the cross-talk weights between PVL cell and EC clusters in IBC_Tumor and IBC_Skin. The band width is proportional to the interaction number. D, Dot plots showing PDGF pathway ligand–receptor interactions between tip cells and PVL cell clusters in each group. The dot size represents the P value calculated by the permutation test, colored by interaction strength. E, Dot plots showing ANGPT pathway ligand–receptor interactions between tip cells and EC clusters in each group. The dot size represents the P value calculated by the permutation test, colored by interaction strength. F, Heatmaps displaying the relative enrichment of each PVL cell cluster among IBC_Skin, non-IBC_Skin, and Normal_Skin (top) and between IBC_Skin and IBC_Tumor (bottom), as calculated by the Ro/e score. Chi-square test; ****, P < 0.0001. G, Representative mIF images for validation of the location of ECs (CD31+), immature PVL cells (MCAM+ CD36+, white arrowheads), and differentiated PVL cells (αSMA+ NG2+ MCAM+, red arrowheads) in the affected skin of patients with IBC. Scale bars, 20 μm. H, Representative mIF images for validation of the location of ECs (CD31+) and differentiated PVL cells (αSMA+ NG2+ MCAM+, red arrowheads) in IBC_Tumors. Scale bars, 20 μm. I, Top, the percentages of immature PVL cells (CD146+ CD36+) in the PVL cell population were compared among the IBC_Skin, non-IBC_Skin, and IBC_Tumor groups. Bottom, the densities of immature PVL cells were compared among IBC_Skin, non-IBC_Skin, and IBC_Tumor. Wilcoxon rank-sum test; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. J, Bar plot showing the enriched GO pathways in PVL_MMP9 and PVL_CCL19 compared with PVL_IGFBP3 and PVL_MYH11.

Close modal

Previous studies have demonstrated that PVL cells are recruited by tip cells to participate in neovascularization (33, 36). We explored the potential interaction network between tip cells and PVL cells and found more frequent cross-talk between these two subpopulations in the affected skin of IBCs than in the skin of non-IBCs with tumors (Fig. 4C; Supplementary Fig. S4E). Notably, no cross-talk was observed between tip cells and PVL cells in normal skin (Supplementary Fig. S4E). In particular, stronger communication between tip cells and PVL cells via the PDGFB–PDGFRB axis was observed in patients with IBC than in patients with non-IBC (Fig. 4D; refs. 33, 36). Interestingly, we observed few interactions between PVL cells and ECs via ANGPT1–TEK, whereas there were strong interactions via ANGPT2–TEK between tip cells and other EC subsets (Fig. 4E; Supplementary Fig. S4F). Angiopoietin-2, a protein encoded by ANGPT2, is largely expressed by activated ECs and antagonizes ANGPT1, which encodes angiopoietin-1, and destabilizes blood vessels by promoting pericyte detachment (37–39).

We next analyzed the distribution of different PVL cells in IBCs and non-IBCs. Notably, compared with non-IBC and normal skin, IBC skin was enriched in PVL_MMP9, and both the PVL_MMP9 and PVL_CCL9 clusters were enriched compared with those in IBC tumors (Fig. 4F; Supplementary Fig. S4G). To study the spatial organization of PVL cells in tumors and skin, we stained CD31 to mark endothelial linings, αSMA, NG2, and MCAM as the markers of PVL cells, and CD36 to identify the PVL_MMP9 and PVL_CCL19 subclusters. mIF staining verified the existence and location of these PVL cells via the coexpression of CD31, MCAM, and CD36 and confirmed their significant accumulation in the affected skin of patients with IBC (Fig. 4GI; Supplementary Figs. S4H and S4I). We further observed that immature PVL cells (MCAM+CD36+) in the affected skin of patients with IBC disseminated throughout the stroma independent of blood vessels (Fig. 4G). Previous studies have demonstrated that CD36 can distinguish immature PVL cells from differentiated PVL cells, whereas elevated RGS5 expression promotes an immature pericyte phenotype, which can result in heightened tumor hypoxia and vessel leakage (12, 34). Notably, RGS5 and CD36 exhibited higher expression levels in PVL_MMP9 and PVL_CCL19 than in PVL_GPFBP3 and PVL_MYH11 (Supplementary Fig. S4C), implying that PVL_MMP9 and PVL_CCL19 are immature and possibly involved in the formation of leaky blood vessels.

To investigate the biological functions of PVL_MMP9 and PVL_CCL19, we performed pathway enrichment analysis. The results showed that pathways related to extracellular matrix (ECM) organization, collagen fibril organization, and cell adhesion were upregulated in the PVL_MMP9 cells. However, in response to hypoxia, Rho protein signaling transduction and the positive regulation of cell migration were elevated in PVL_CCL19 cells (Fig. 4J). These findings are consistent with a previous study showing that the loss of vascular smooth muscle cell markers in PVL cells is abundant in areas other than the vasculature at premetastatic sites, which synthesizes fibronectin-containing ECM to support the metastatic behavior of disseminated malignant cells (40).

Overall, these observations suggest that PVL_MMP9 and PVL_CCL19, which are enriched in the affected skin of IBC patients, promote tumor cell metastasis, in addition to being involved in angiogenesis.

Malignant cells in the affected skin of patients with IBC are more invasive and prone to metastasis

On the basis of the enrichment of immature PVL cells in the affected skin of patients with IBC and their potential role in promoting metastasis, we next explored the connection between the PVL and malignant cells in IBC. First, we analyzed the differences in the molecular and genomic landscapes of malignant cells between IBC and non-IBC tissues in our scRNA-seq data. Genes related to tumor cell proliferation and metastasis, such as DLK1 (41), FGF19 (42), CDK4 (43), LASP1 (44), and CCND1 (43), were highly expressed in IBC-malignant cells (Supplementary Fig. S5A). Compared with those in non-IBC cells, pathways associated with the cell cycle, Wnt signaling, and cell migration were enriched in IBC-malignant cells (Supplementary Fig. S5B), which is consistent with previous findings (45, 46). We next explored how expression states varied among different malignant cells in both patients with IBC and non-IBC. Consensus non-negative matrix factorization (cNMF) was performed to investigate coherent sets of genes that were preferentially coexpressed by subsets of malignant cells. Five expression programs, including E2F targets, TNFα signaling via NF-κB, oxidative phosphorylation, coagulation, and epithelial–mesenchymal transition (EMT), were detected in both IBC and non-IBC malignant cells (Fig. 5A; Supplementary Table S14), but no specific program unique to IBC malignant cells was identified (Fig. 5B). In addition, we analyzed the WES data of 15 IBC and 38 non-IBC tumors, 33 of which were non-IBC tumors obtained from TCGA database, to characterize the mutational landscape of the genome. Consistent with previous reports (43, 47), TP53 (53% and 55%) was the most highly mutated gene in both IBC and non-IBC tumors (Supplementary Figs. S5C and S5D), and no obviously different mutants were characterized.

Figure 5.

Malignant cells in the affected skin of patients with IBC are more invasive and prone to metastasis. A, Heatmap depicting pairwise correlations of intratumoral programs derived from IBC_Tumor and non-IBC_Tumor tissues. Clustering revealed five coherent expression programs across tumor samples. B, The proportions of IBC_Tumor and non-IBC_Tumor enriched in the five programs. C, Bar plot showing the enriched GO pathways in malignant cells in IBC_Skin compared with malignant cells in IBC_Tumor. D, Violin plot showing comparisons of the EMT scores of malignant cells in IBC_Skin and IBC_Tumor. Wilcoxon rank-sum test; ***, P < 0.001. E, Dot plot showing the significantly upregulated signaling ligand–receptor interactions between PVL cells and malignant cell clusters in IBC_Tumor and IBC_Skin. The dot size represents the P value calculated by the permutation test, colored by interaction strength. F, Dot plot showing ANGPTL4 signaling ligand–receptor interactions between malignant cells and EC clusters in IBC_Tumor and IBC_Skin. The dot size represents the P value calculated by the permutation test, colored by interaction strength.

Figure 5.

Malignant cells in the affected skin of patients with IBC are more invasive and prone to metastasis. A, Heatmap depicting pairwise correlations of intratumoral programs derived from IBC_Tumor and non-IBC_Tumor tissues. Clustering revealed five coherent expression programs across tumor samples. B, The proportions of IBC_Tumor and non-IBC_Tumor enriched in the five programs. C, Bar plot showing the enriched GO pathways in malignant cells in IBC_Skin compared with malignant cells in IBC_Tumor. D, Violin plot showing comparisons of the EMT scores of malignant cells in IBC_Skin and IBC_Tumor. Wilcoxon rank-sum test; ***, P < 0.001. E, Dot plot showing the significantly upregulated signaling ligand–receptor interactions between PVL cells and malignant cell clusters in IBC_Tumor and IBC_Skin. The dot size represents the P value calculated by the permutation test, colored by interaction strength. F, Dot plot showing ANGPTL4 signaling ligand–receptor interactions between malignant cells and EC clusters in IBC_Tumor and IBC_Skin. The dot size represents the P value calculated by the permutation test, colored by interaction strength.

Close modal

Next, we compared the transcriptomic differences of malignant cells between tumors and the affected skin and found that malignant cells in the affected skin of patients with IBC highly expressed genes involved in pathways regulating cell migration, actin filament organization, and the positive regulation of EMT (Fig. 5C). Furthermore, malignant cells in the affected skin of patients with IBC had a higher EMT score than did intratumoral malignant cells (Fig. 5D; Supplementary Table S15). L–R analyses revealed strong communication between PVL_MMP9/PVL_CCL19 and malignant cells via the interactions of TNFSF12-TNFRSF12A (48), TGFB-TGFBR1/TGFBR2 (49), and MDK-SDC1/4/NCL/(ITGA6+ITGB1; Fig. 5E; ref. 50), suggesting the role of PVL cells in enhancing malignant cell metastasis in the affected skin of patients with IBC. Vasculature leakiness plays a key role in metastatic spread and affects malignant cell intravasation and extravasation, which allows directional migration of malignant cells across the disrupted endothelium. We observed robust cross-talk between malignant cells and EC subsets via the ANGPTL4-CLDN5/CDH5/ITGA5/ITGB1 axes in the affected skin of only patients with IBC (Fig. 5F). Tumor cell-derived ANGPTL4 can regulate vascular junction integrity through integrin signaling and disruption of intercellular VE-cadherin and claudin-5 clusters (51, 52). Overall, our data demonstrate that malignant cells in the affected skin of patients with IBC have a more aggressive phenotype, which may be associated with the enrichment of the PVL_MMP9/PVL_CCL19 subsets, compared with that in patients with IBC, thus explaining, in part, why IBCs with distinctive cutaneous presentations have a strong ability to metastasize.

PTN promotes EC motility and sprouting via binding with NRP1 in a VEGFA-independent manner

On the basis of our finding that PTN may be a critical mediator in promoting the unique and aggressive phenotype of IBC, we next sought to confirm whether PTN could induce the migration of ECs, vessel sprouting, and tube formation. Dose–response experiments revealed that the addition of recombinant PTN protein promoted HUVEC migration and tube formation in vitro (Fig. 6A and B). We also conducted a HUVEC sprouting assay in three-dimensional (3D) culture. Consistent with previous results, the average length and number of sprouts increased upon PTN administration (Fig. 6C). We investigated whether the function of PTN in angiogenesis is dependent on its interaction with NRP1, and we observed that knocking down NRP1 dramatically suppressed the migratory, tube formation, and sprouting abilities of HUVECs induced by recombinant PTN protein (Fig. 6DH). The expression of other potential PTN receptors was detected using RT-qPCR technology. The result demonstrated that the expression levels of SDC3 and NCL were significantly lower than that of NRP1 (Supplementary Fig. S6A), indicating that PTN primarily exerts its function through NRP1.

Figure 6.

PTN promotes EC motility and sprouting via binding with NRP1 in a VEGFA-independent manner. A, Transwell assays were used to detect the migration of HUVECs treated with recombinant PTN protein at different concentrations and untreated HUVECs. Representative images (left) and quantification (right) are shown. Unpaired t test; ****, P < 0.0001. B, Tube formation assays were used to detect angiogenesis in HUVECs treated with recombinant PTN protein at different concentrations and in untreated HUVECs. Unpaired t test; *, P < 0.05; **, P < 0.01; ****, P < 0.0001. C, HUVEC sprouting 3D cultures were treated with recombinant PTN protein at different concentrations and stained with phalloidin (green) and DAPI (blue). Scale bars, 50 μm. Unpaired t test; *, P < 0.05; **, P < 0.01; ****, P < 0.0001. D, RT-qPCR confirmation of NRP1 knockdown in HUVECs using two independent siRNAs. Unpaired t test; *, P < 0.05. E, Left, Western blotting confirmation of the effects of NRP1 knockdown on HUVECs using two siRNAs. Right, the intensities of the Western blot bands were quantified. Unpaired t test; *, P < 0.05. F, Transwell assays were used to detect HUVEC migration upon NRP1 knockdown. Unpaired t test; ****, P < 0.0001. G, Tube formation assays were used to detect HUVEC angiogenesis upon NRP1 knockdown. Unpaired t test; ***, P < 0.001; ****, P < 0.0001. H, HUVEC sprouting 3D cultures were used to detect the ability of HUVECs to sprout upon NRP1 knockdown, and the cells were stained with phalloidin (green) and DAPI (blue). Scale bars, 50 μm. Unpaired t test; ****, P < 0.0001. I, Transwell assays showing that bevacizumab failed to inhibit the migration of HUVECs treated with recombinant PTN protein. Unpaired t test; **, P < 0.01; ****, P < 0.0001. J, Tube formation assays showed that bevacizumab failed to inhibit the tube formation of HUVECs induced by recombinant PTN protein. Unpaired t test; ***, P < 0.001; ****, P < 0.0001. K, Comparison of the integrated optical density (IOD) of PTNs quantified by IHC between responders (n = 18) and nonresponders (n = 9). Wilcoxon rank-sum test; *, P < 0.05. ns, nonsignificant.

Figure 6.

PTN promotes EC motility and sprouting via binding with NRP1 in a VEGFA-independent manner. A, Transwell assays were used to detect the migration of HUVECs treated with recombinant PTN protein at different concentrations and untreated HUVECs. Representative images (left) and quantification (right) are shown. Unpaired t test; ****, P < 0.0001. B, Tube formation assays were used to detect angiogenesis in HUVECs treated with recombinant PTN protein at different concentrations and in untreated HUVECs. Unpaired t test; *, P < 0.05; **, P < 0.01; ****, P < 0.0001. C, HUVEC sprouting 3D cultures were treated with recombinant PTN protein at different concentrations and stained with phalloidin (green) and DAPI (blue). Scale bars, 50 μm. Unpaired t test; *, P < 0.05; **, P < 0.01; ****, P < 0.0001. D, RT-qPCR confirmation of NRP1 knockdown in HUVECs using two independent siRNAs. Unpaired t test; *, P < 0.05. E, Left, Western blotting confirmation of the effects of NRP1 knockdown on HUVECs using two siRNAs. Right, the intensities of the Western blot bands were quantified. Unpaired t test; *, P < 0.05. F, Transwell assays were used to detect HUVEC migration upon NRP1 knockdown. Unpaired t test; ****, P < 0.0001. G, Tube formation assays were used to detect HUVEC angiogenesis upon NRP1 knockdown. Unpaired t test; ***, P < 0.001; ****, P < 0.0001. H, HUVEC sprouting 3D cultures were used to detect the ability of HUVECs to sprout upon NRP1 knockdown, and the cells were stained with phalloidin (green) and DAPI (blue). Scale bars, 50 μm. Unpaired t test; ****, P < 0.0001. I, Transwell assays showing that bevacizumab failed to inhibit the migration of HUVECs treated with recombinant PTN protein. Unpaired t test; **, P < 0.01; ****, P < 0.0001. J, Tube formation assays showed that bevacizumab failed to inhibit the tube formation of HUVECs induced by recombinant PTN protein. Unpaired t test; ***, P < 0.001; ****, P < 0.0001. K, Comparison of the integrated optical density (IOD) of PTNs quantified by IHC between responders (n = 18) and nonresponders (n = 9). Wilcoxon rank-sum test; *, P < 0.05. ns, nonsignificant.

Close modal

To further investigate whether the function of PTN in angiogenesis depends on VEGFA, we performed transcriptome sequencing on HUVECs before and after stimulation with various doses of recombinant PTN protein. The results demonstrated that there was no significant change in the expression of VEGFA or VEGF-related signaling pathway (Supplementary Fig. S6B; Supplementary Table S11). Similar results were also observed in the RT-qPCR and ELISA analyses (Supplementary Figs. S6C and S6D). Functional assays revealed that PTN-induced migration and tube formation of HUVECs were not affected by bevacizumab, a recombinant humanized monoclonal antibody for VEGFA blockage (Fig. 6I and J), or vandetanib, a VEGFR2 inhibitor (Supplementary Fig. S6E). Previous studies have confirmed that VEGFA promotes angiogenesis mainly by binding to VEGFR2 and activating downstream pathways (53, 54). These findings indicate that PTN exerts its proangiogenic effect independent of VEGFA and its associated signaling pathway.

We investigated whether VEGFA competes with PTN for binding to the NRP1 receptor, and found that depletion of NRP1 did not suppress the HUVEC migratory and tube formation abilities induced by the recombinant VEGFA-165 protein, whereas vandetanib significantly suppressed the phenotypes induced by VEGFA-165 (Supplementary Figs. S6F and S6G). Specifically, vandetanib did not further suppress migration or tube formation in NRP1-deficient HUVECs (Supplementary Figs. S6F and S6G). These results suggest that VEGFA promotes angiogenesis by binding to VEGFR2 rather than competing with PTN for binding to the NRP1 receptor.

Finally, we examined whether PTN is related to the efficacy of systemic therapy in patients with IBC. We observed that patients with IBC with high PTN expression were more prone to treatment resistance (including chemotherapy and targeted therapies) than were those with low PTN expression (Fig. 6K). Furthermore, patients with either HER2-positive (Supplementary Fig. S6H) or triple-negative breast cancer (TNBC) that had high PTN expression (Supplementary Fig. S6I) exhibited worse overall survival (OS) and disease-free survival (DFS) than did those with low PTN expression. These results indicate that PTN may be closely related to prognosis in highly aggressive breast cancers.

Collectively, these in vitro findings suggest that the PTN–NRP1 axis participates in the regulation of angiogenesis. The integration of clinical data and TCGA analysis also highlights the critical role of PTN in IBC progression, possibly through proangiogenic functions.

Previous genomic and transcriptomic studies based on bulk analyses have been used to discover novel biological targets in IBC, yet the underlying mechanisms remain elusive (4, 9, 10, 55). Current single-cell transcriptomic studies have emerged as powerful tools for high-dimensional, comprehensive characterization of the tumor microenvironment (56). To our knowledge, this is the first report delineating the landscape of IBC tumors and the affected skin microenvironment at the single-cell level. With this approach, we determined that the PTN+LP subpopulation is the crucial contributor to the distinct clinical presentation and aggressive nature of IBC. PTN+ LP cells interact with ECs via the PTN–NRP1 axis to promote angiogenesis, which might be independent of the VEGFA and VEGF signaling pathways, resulting in tip cell enrichment in both tumors and the affected skin of patients with IBC (Fig. 7). During this process, immature PVL_MMP9 and PVL_CCL19 cells that accumulate in the affected skin of patients with IBC may promote tumor metastasis.

Figure 7.

Schematic diagram showing how PTN+LP cells promote angiogenesis and metastasis in IBC. PTN+ LP cells directly interact with ECs via the PTN–NRP1 axis to promote angiogenesis, resulting in tip cell enrichment in both the tumors and affected skin of patients with IBC and potentially contributing to the metastasis of malignant cells in the affected skin. Targeting the PTN–NRP1 axis might be a potential therapy for IBC.

Figure 7.

Schematic diagram showing how PTN+LP cells promote angiogenesis and metastasis in IBC. PTN+ LP cells directly interact with ECs via the PTN–NRP1 axis to promote angiogenesis, resulting in tip cell enrichment in both the tumors and affected skin of patients with IBC and potentially contributing to the metastasis of malignant cells in the affected skin. Targeting the PTN–NRP1 axis might be a potential therapy for IBC.

Close modal

Previous studies have shown that LP cells in normal mammary glands highly express genes associated with milk production (LTF), secretory molecules (SAA1 and SAA2), and epithelial keratin molecules (KRT8, KRT18, KRT5, and KRT14), suggesting that LP cells can further differentiate into hormone-sensitive mature luminal cells and secretory mature luminal cells in normal breast tissues (19, 22). Although it has been reported that LP cells may be the cell of origin in the basal-like subtype of breast cancer (57), how LP cells contribute to the microenvironment remains unclear (15, 58). Strikingly, our results revealed that in addition to the expression of LP cell lineage genes (KRT8, KRT18, KRT5) and stem cell genes (ITGA6, KIT), LP cells in IBC exhibited the highest expression level of PTN, which has not been reported previously. Several risk factors for IBC, including younger age at first birth, not breastfeeding or breastfeeding for a short duration, which may lead to abnormal changes in normal breast tissue prior to tumor development, have been identified through epidemiologic studies (59, 60). Pregnancy induces a transient 11-fold increase in the number of mammary stem cells (CD44+CD29+) in mice (61). Furthermore, postlactational involution is a multistage process that involves apoptosis and tissue remodeling, facilitating the regression of the mammary gland epithelium to its non-lactating state. Therefore, researchers have proposed that abrupt or premature breast involution, such as limited breastfeeding, leads to the aberrant accumulation of mammary stem cells (62). A significantly enriched post-lactational mammary gland involution signature was identified in IBC rather than non-IBC samples by analyzing the gene expression profiles of mammary glands derived from force-weaned mice during their peak lactation period (59). In addition, a study identified a subpopulation of CD44+CD49f+CD133/2+ stem cells in adjacent normal IBC tissues but not in non-IBC tissues (14). Notably, we also observed enrichment of ACTA2+ Myo cells in IBC tumors. However, these cells displayed high expression of genes associated with myoepithelial cell function (e.g., ACTA2, TGLN, MYL9; ref. 22), suggesting that these cells may not have a distinct biological function in the IBC microenvironment. Collectively, we demonstrated that the accumulation and distinct changes in LP cells in the mammary glands may be key factors involved in inducing IBC formation.

Our analysis expands upon those previous findings, where we revealed that the expression level of PTN was significantly higher in IBC-LP cells than in other nonmalignant epithelial cells, malignant cells, non-IBC-LP cells, and normal breast-LP cells. The PTN gene encodes PTN, a cytokine that plays multifunctional roles in various biological processes, including neuronal development (63), adipocyte differentiation (64), delaying mammary epithelial cell differentiation (65), and angiogenesis (66). We demonstrated that PTN can be secreted in patients with IBC and is involved in angiogenesis. The TCGA database showed lower PTN expression in breast cancer tissues than in adjacent normal breast tissues (67). Furthermore, Chang and colleagues confirmed that PTN is not an oncogene (68). In our study, the expression of PTN in malignant cells was extremely low, and mutations in PTN were not detected by WES. Thus, previous studies and our data suggest that PTN does not originate from malignant cells in breast cancer. We found that patients with HER2 overexpression and TNBC with high PTN expression had worse OS and DFS than those with low PTN expression in the TCGA cohort, which is of great clinical interest. A recent study revealed that PTN is associated with the metastatic niche in metastatic breast cancer (69). Therefore, we speculate that these findings might have far-reaching implications for highly aggressive breast cancer. However, how PTN expression is upregulated in the LP cells of patients with IBC remains unclear. Further mechanistic studies are required to elucidate this regulatory mechanism and intervention experiments are needed to guide future therapeutic development.

Specialized ECs, known as tip cells, spearhead the growth of vessels during sprouting angiogenesis (70) and are not detectable in normal tissue (26). Our data showed that tip cells were enriched in tumors and the affected skin of patients with IBC, indicating hyperactivated neoangiogenesis in IBC. Previous studies have also confirmed that IBC tumors exhibit intense angiogenic activity and that their microvessel density is higher than that of non-IBC tumors, suggesting that anti-angiogenic drugs could be effective in treating IBC (71, 72). However, the efficacy of anti-VEGFA therapy is limited in patients with IBC. We confirmed the important role of the PTN–NRP1 axis in the regulation of angiogenesis. Furthermore, this angiogenesis–promoting interaction of the PTN–NRP1 axis may be independent of the VEGFA and VEGF signaling pathways in IBC. Our findings suggested that the PTN–NRP1 axis is the key proangiogenic factor in IBC and could explain why patients with IBC cannot benefit from the combination of anti-VEGFA agents with chemotherapy (32). Therefore, we propose that high PTN expression in LP cells may not cause breast cancer but may promote angiogenesis, providing a novel therapeutic target for IBC treatment. However, the downstream regulatory mechanisms of the PTN–NRP1 axis need to be further explored. On the other hand, currently, there are no potent drugs that specifically target PTN. Moreover, it is speculated that targeted therapy against PTN may damage the central nervous system (CNS), leading to cognitive side effects or affecting patients’ tissue repair functions. Thus, developing novel drug delivery systems or designing agents specifically optimized for blood–brain barrier penetration may help refine the specificity of PTN-targeting agents suitable for use against IBC while minimizing central nervous system side effects.

A previous study revealed that PVL cells that lack the expression of contractile genes (such as MYH11, ACTA2, and TAGLN) can establish a prometastatic fibronectin-rich environment away from the vasculature, and these PVL cells can acquire a proliferative, migrating, and extracellular matrix-modulating phenotype (40). Furthermore, co-injection of mesenchymal stem cells and IBC tumor cells in preclinical models could induce skin erythema and the formation of subcutaneous tumor emboli, reflecting the indispensable role of mesenchymal stem cells in the formation of the IBC clinical phenotype (13). Several studies have confirmed that PVL cells mainly evolve from mesenchymal stem cells (73, 74). In this study, we first described that immature PVL cells disseminate throughout the stroma independent of blood vessels and may play a key role in regulating the metastatic microenvironment in the affected skin of patients with IBC to support malignant cell metastasis. Our work provides a preliminary overview of the microenvironment involved in skin redness and edema, and the underlying mechanisms of prometastasis will be further investigated in our future studies.

In conclusion, our results identify a unique, nonmalignant epithelial cell subpopulation, PTN+ LP cells, that not only promotes angiogenesis via the PTN–NRP1 axis in IBC but also plays a pivotal role in building the prometastatic tumor microenvironment in IBC affected skin. These findings offer comprehensive, in-depth insights into the pathogenesis and progression of IBC that might lead to novel therapeutic targets in IBC.

L.E. Stevens reports personal fees from AstraZeneca outside the submitted work. F. Lynce reports grants and personal fees from AstraZeneca and Daiichi Sankyo; personal fees from Pfizer and Eli Lilly; grants from Gilead, Zentalis, Merck, and Ideya outside the submitted work. No disclosures were reported by the other authors.

M. Zhang: Conceptualization, resources, software, formal analysis, validation, investigation, visualization, methodology, writing–original draft, project administration. K. Zhou: Software, formal analysis, investigation, methodology, writing–original draft. Z. Wang: Validation, investigation, methodology, writing–original draft. T. Liu: Validation, writing–original draft. L.E. Stevens: Software, supervision, methodology, writing–review and editing. F. Lynce: Supervision, validation, methodology, writing–review and editing. W.Y. Chen: Supervision, methodology, writing–review and editing. S. Peng: Software, supervision, funding acquisition, methodology. Y. Xie: Software, validation, methodology. D. Zhai: Resources, data curation, validation. Q. Chen: Resources, data curation. Y. Shi: Resources, data curation. H. Shi: Resources, data curation. Z. Yuan: Resources, data curation. X. Li: Resources, data curation. J. Xu: Resources, data curation. Z. Cai: Resources, data curation. J. Guo: Conceptualization, supervision, validation, investigation, visualization, methodology, project administration, writing–review and editing. N. Shao: Conceptualization, resources, data curation, supervision, methodology, writing–original draft, project administration. Y. Lin: Conceptualization, resources, data curation, supervision, funding acquisition, validation, writing-original draft, project administration, writing–review and editing.

This study was supported by the National Key Research and Development Program (2021YFE0206300), the National Natural Science Foundation of China (82372781), and the Beijing Science and Technology Innovation Medical Development Foundation (KC2021-JX-0044–11).

Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).

1.
Hance
KW
,
Anderson
WF
,
Devesa
SS
,
Young
HA
,
Levine
PH
.
Trends in inflammatory breast carcinoma incidence and survival: the surveillance, epidemiology, and end results program at the National Cancer Institute
.
J Natl Cancer Inst
2005
;
97
:
966
75
.
2.
Low
JA
,
Berman
AW
,
Steinberg
SM
,
Danforth
DN
,
Lippman
ME
,
Swain
SM
.
Long-term follow-up for locally advanced and inflammatory breast cancer patients treated with multimodality therapy
.
J Clin Oncol
2004
;
22
:
4067
74
.
3.
Valente
SA
,
ElSherif
A
,
Tran
J
,
Tu
C
,
Kruse
M
.
Inflammatory breast cancer and the importance of skin punch biopsy
.
Am J Surg
2023
;
225
:
549
52
.
4.
Van Laere
SJ
,
Ueno
NT
,
Finetti
P
,
Vermeulen
P
,
Lucci
A
,
Robertson
FM
, et al
.
Uncovering the molecular secrets of inflammatory breast cancer biology: an integrated analysis of three distinct affymetrix gene expression datasets
.
Clin Cancer Res
2013
;
19
:
4685
96
.
5.
Kertmen
N
,
Babacan
T
,
Keskin
O
,
Solak
M
,
Sarici
F
,
Akin
S
, et al
.
Molecular subtypes in patients with inflammatory breast cancer; a single center experience
.
J buon
2015
;
20
:
35
9
.
6.
Bonnier
P
,
Charpin
C
,
Lejeune
C
,
Romain
S
,
Tubiana
N
,
Beedassy
B
, et al
.
Inflammatory carcinomas of the breast: a clinical, pathological, or a clinical and pathological definition?
Int J Cancer
1995
;
62
:
382
5
.
7.
Manfrin
E
,
Remo
A
,
Pancione
M
,
Cannizzaro
C
,
Falsirollo
F
,
Pollini
GP
, et al
.
Comparison between invasive breast cancer with extensive peritumoral vascular invasion and inflammatory breast carcinoma: a clinicopathologic study of 161 cases
.
Am J Clin Pathol
2014
;
142
:
299
306
.
8.
Masuda
H
,
Brewer
TM
,
Liu
DD
,
Iwamoto
T
,
Shen
Y
,
Hsu
L
, et al
.
Long-term treatment efficacy in primary inflammatory breast cancer by hormonal receptor- and HER2-defined subtypes
.
Ann Oncol
2014
;
25
:
384
91
.
9.
Masuda
H
,
Baggerly
KA
,
Wang
Y
,
Iwamoto
T
,
Brewer
T
,
Pusztai
L
, et al
.
Comparison of molecular subtype distribution in triple-negative inflammatory and non-inflammatory breast cancers
.
Breast Cancer Res
2013
;
15
:
R112
.
10.
Li
X
,
Kumar
S
,
Harmanci
A
,
Li
S
,
Kitchen
RR
,
Zhang
Y
, et al
.
Whole-genome sequencing of phenotypically distinct inflammatory breast cancers reveals similar genomic alterations to non-inflammatory breast cancers
.
Genome Med
2021
;
13
:
70
.
11.
Wu
SZ
,
Al-Eryani
G
,
Roden
DL
,
Junankar
S
,
Harvey
K
,
Andersson
A
, et al
.
A single-cell and spatially resolved atlas of human breast cancers
.
Nat Genet
2021
;
53
:
1334
47
.
12.
Wu
SZ
,
Roden
DL
,
Wang
C
,
Holliday
H
,
Harvey
K
,
Cazet
AS
, et al
.
Stromal cell diversity associated with immune evasion in human triple-negative breast cancer
.
Embo j
2020
;
39
:
e104063
.
13.
Lacerda
L
,
Debeb
BG
,
Smith
D
,
Larson
R
,
Solley
T
,
Xu
W
, et al
.
Mesenchymal stem cells mediate the clinical phenotype of inflammatory breast cancer in a preclinical model
.
Breast Cancer Res
2015
;
17
:
42
.
14.
Reddy
JP
,
Atkinson
RL
,
Larson
R
,
Burks
JK
,
Smith
D
,
Debeb
BG
, et al
.
Mammary stem cell and macrophage markers are enriched in normal tissue adjacent to inflammatory breast cancer
.
Breast Cancer Res Treat
2018
;
171
:
283
93
.
15.
Koren
S
,
Bentires-Alj
M
.
Breast tumor heterogeneity: source of fitness, hurdle for therapy
.
Mol Cell
2015
;
60
:
537
46
.
16.
McGinnis
CS
,
Murrow
LM
,
Gartner
ZJ
.
DoubletFinder
.:
Doublet detection in single-cell RNA sequencing data using artificial nearest neighbors
.
Cell Syst
2019
;
8
:
329
37
.
17.
Haghverdi
L
,
Lun
ATL
,
Morgan
MD
,
Marioni
JC
.
Batch effects in single-cell RNA-sequencing data are corrected by matching mutual nearest neighbors
.
Nat Biotechnol
2018
;
36
:
421
7
.
18.
Nakatsu
MN
,
Sainson
RC
,
Aoto
JN
,
Taylor
KL
,
Aitkenhead
M
,
Pérez-del-Pulgar
S
, et al
.
Angiogenic sprouting and capillary lumen formation modeled by human umbilical vein endothelial cells (HUVEC) in fibrin gels: the role of fibroblasts and Angiopoietin-1
.
Microvasc Res
2003
;
66
:
102
12
.
19.
Pal
B
,
Chen
Y
,
Vaillant
F
,
Capaldo
BD
,
Joyce
R
,
Song
X
, et al
.
A single-cell RNA expression atlas of normal, preneoplastic and tumorigenic states in the human breast
.
Embo j
2021
;
40
:
e107333
.
20.
Bassez
A
,
Vos
H
,
Van Dyck
L
,
Floris
G
,
Arijs
I
,
Desmedt
C
, et al
.
A single-cell map of intratumoral changes during anti-PD1 treatment of patients with breast cancer
.
Nat Med
2021
;
27
:
820
32
.
21.
Puram
SV
,
Tirosh
I
,
Parikh
AS
,
Patel
AP
,
Yizhak
K
,
Gillespie
S
, et al
.
Single-cell transcriptomic analysis of primary and metastatic tumor ecosystems in head and neck cancer
.
Cell
2017
;
171
:
1611
24
.
22.
Nguyen
QH
,
Pervolarakis
N
,
Blake
K
,
Ma
D
,
Davis
RT
,
James
N
, et al
.
Profiling human breast epithelial cells using single cell RNA sequencing identifies cell diversity
.
Nat Commun
2018
;
9
:
2028
.
23.
Li
YS
,
Milner
PG
,
Chauhan
AK
,
Watson
MA
,
Hoffman
RM
,
Kodner
CM
, et al
.
Cloning and expression of a developmentally regulated protein that induces mitogenic and neurite outgrowth activity
.
Science
1990
;
250
:
1690
4
.
24.
Guo
X
,
Zhang
Y
,
Zheng
L
,
Zheng
C
,
Song
J
,
Zhang
Q
, et al
.
Global characterization of T cells in non-small-cell lung cancer by single-cell sequencing
.
Nat Med
2018
;
24
:
978
85
.
25.
Zhang
L
,
Li
Z
,
Skrzypczynska
KM
,
Fang
Q
,
Zhang
W
,
O'Brien
SA
, et al
.
Single-cell analyses inform mechanisms of myeloid-targeted therapies in colon cancer
.
Cell
2020
;
181
:
442
59
.
26.
Goveia
J
,
Rohlenova
K
,
Taverna
F
,
Treps
L
,
Conradi
LC
,
Pircher
A
, et al
.
An integrated gene expression landscape profiling approach to identify lung tumor endothelial cell heterogeneity and angiogenic candidates
.
Cancer Cell
2020
;
37
:
21
36
.
27.
Zhao
Q
,
Eichten
A
,
Parveen
A
,
Adler
C
,
Huang
Y
,
Wang
W
, et al
.
Single-cell transcriptome analyses reveal endothelial cell heterogeneity in tumors and changes following antiangiogenic treatment
.
Cancer Res
2018
;
78
:
2370
82
.
28.
Seano
G
,
Chiaverina
G
,
Gagliardi
PA
,
di Blasio
L
,
Puliafito
A
,
Bouvard
C
, et al
.
Endothelial podosome rosettes regulate vascular branching in tumour angiogenesis
.
Nat Cell Biol
2014
;
16
:
931
41
,
1
8
.
29.
Papadimitriou
E
,
Pantazaka
E
,
Castana
P
,
Tsalios
T
,
Polyzos
A
,
Beis
D
.
Pleiotrophin and its receptor protein tyrosine phosphatase beta/zeta as regulators of angiogenesis and cancer
.
Biochim Biophys Acta
2016
;
1866
:
252
65
.
30.
Elahouel
R
,
Blanc
C
,
Carpentier
G
,
Frechault
S
,
Cascone
I
,
Destouches
D
, et al
.
Pleiotrophin exerts its migration and invasion effect through the neuropilin-1 pathway
.
Neoplasia
2015
;
17
:
613
24
.
31.
Ferrara
N
,
Gerber
HP
,
LeCouter
J
.
The biology of VEGF and its receptors
.
Nat Med
2003
;
9
:
669
76
.
32.
Bertucci
F
,
Fekih
M
,
Autret
A
,
Petit
T
,
Dalenc
F
,
Levy
C
, et al
.
Bevacizumab plus neoadjuvant chemotherapy in patients with HER2-negative inflammatory breast cancer (BEVERLY-1): a multicentre, single-arm, phase 2 study
.
Lancet Oncol
2016
;
17
:
600
11
.
33.
Armulik
A
,
Genové
G
,
Betsholtz
C
.
Pericytes: developmental, physiological, and pathological perspectives, problems, and promises
.
Dev Cell
2011
;
21
:
193
215
.
34.
Hamzah
J
,
Jugold
M
,
Kiessling
F
,
Rigby
P
,
Manzur
M
,
Marti
HH
, et al
.
Vascular normalization in Rgs5-deficient tumours promotes immune destruction
.
Nature
2008
;
453
:
410
4
.
35.
Kim
J
,
de Sampaio
PC
,
Lundy
DM
,
Peng
Q
,
Evans
KW
,
Sugimoto
H
, et al
.
Heterogeneous perivascular cell coverage affects breast cancer metastasis and response to chemotherapy
.
JCI Insight
2016
;
1
:
e90733
.
36.
Gerhardt
H
,
Betsholtz
C
.
Endothelial-pericyte interactions in angiogenesis
.
Cell Tissue Res
2003
;
314
:
15
23
.
37.
Srivastava
K
,
Hu
J
,
Korn
C
,
Savant
S
,
Teichert
M
,
Kapel
SS
, et al
.
Postsurgical adjuvant tumor therapy by combining anti-angiopoietin-2 and metronomic chemotherapy limits metastatic growth
.
Cancer Cell
2014
;
26
:
880
95
.
38.
Park
JS
,
Kim
IK
,
Han
S
,
Park
I
,
Kim
C
,
Bae
J
, et al
.
Normalization of tumor vessels by Tie2 activation and ang2 inhibition enhances drug delivery and produces a favorable tumor microenvironment
.
Cancer Cell
2016
;
30
:
953
67
.
39.
Augustin
HG
,
Koh
GY
,
Thurston
G
,
Alitalo
K
.
Control of vascular morphogenesis and homeostasis through the angiopoietin-Tie system
.
Nat Rev Mol Cell Biol
2009
;
10
:
165
77
.
40.
Murgai
M
,
Ju
W
,
Eason
M
,
Kline
J
,
Beury
DW
,
Kaczanowska
S
, et al
.
KLF4-dependent perivascular cell plasticity mediates pre-metastatic niche formation and metastasis
.
Nat Med
2017
;
23
:
1176
90
.
41.
Pittaway
JFH
,
Lipsos
C
,
Mariniello
K
,
Guasti
L
.
The role of delta-like non-canonical Notch ligand 1 (DLK1) in cancer
.
Endocr Relat Cancer
2021
;
28
:
R271
r87
.
42.
Zhao
X
,
Xu
F
,
Dominguez
NP
,
Xiong
Y
,
Xiong
Z
,
Peng
H
, et al
.
FGFR4 provides the conduit to facilitate FGF19 signaling in breast cancer progression
.
Mol Carcinog
2018
;
57
:
1616
25
.
43.
Bertucci
F
,
Rypens
C
,
Finetti
P
,
Guille
A
,
Adélaïde
J
,
Monneur
A
, et al
.
NOTCH and DNA repair pathways are more frequently targeted by genomic alterations in inflammatory than in non-inflammatory breast cancers
.
Mol Oncol
2020
;
14
:
504
19
.
44.
Grunewald
TG
,
Kammerer
U
,
Schulze
E
,
Schindler
D
,
Honig
A
,
Zimmer
M
, et al
.
Silencing of LASP-1 influences zyxin localization, inhibits proliferation and reduces migration in breast cancer cells
.
Exp Cell Res
2006
;
312
:
974
82
.
45.
van Golen
KL
,
Wu
ZF
,
Qiao
XT
,
Bao
LW
,
Merajver
SD
.
RhoC GTPase, a novel transforming oncogene for human mammary epithelial cells that partially recapitulates the inflammatory breast cancer phenotype
.
Cancer Res
2000
;
60
:
5832
8
.
46.
Chakraborty
P
,
George
JT
,
Woodward
WA
,
Levine
H
,
Jolly
MK
.
Gene expression profiles of inflammatory breast cancer reveal high heterogeneity across the epithelial-hybrid-mesenchymal spectrum
.
Transl Oncol
2021
;
14
:
101026
.
47.
Ross
JS
,
Ali
SM
,
Wang
K
,
Khaira
D
,
Palma
NA
,
Chmielecki
J
, et al
.
Comprehensive genomic profiling of inflammatory breast cancer cases reveals a high frequency of clinically relevant genomic alterations
.
Breast Cancer Res Treat
2015
;
154
:
155
62
.
48.
Dwyer
BJ
,
Jarman
EJ
,
Gogoi-Tiwari
J
,
Ferreira-Gonzalez
S
,
Boulter
L
,
Guest
RV
, et al
.
TWEAK/Fn14 signalling promotes cholangiocarcinoma niche formation and progression
.
J Hepatol
2021
;
74
:
860
72
.
49.
Tsubakihara
Y
,
Moustakas
A
.
Epithelial-mesenchymal transition and metastasis under the control of transforming growth factor β
.
Int J Mol Sci
2018
;
19
:
3672
.
50.
Filippou
PS
,
Karagiannis
GS
, and
Constantinidou
A
.
Midkine
. (
MDK) growth factor: a key player in cancer progression and a promising therapeutic target
.
Oncogene
2020
;
39
:
2040
54
.
51.
Huang
RL
,
Teo
Z
,
Chong
HC
,
Zhu
P
,
Tan
MJ
,
Tan
CK
, et al
.
ANGPTL4 modulates vascular junction integrity by integrin signaling and disruption of intercellular VE-cadherin and claudin-5 clusters
.
Blood
2011
;
118
:
3990
4002
.
52.
Padua
D
,
Zhang
XH
,
Wang
Q
,
Nadal
C
,
Gerald
WL
,
Gomis
RR
, et al
.
TGFbeta primes breast tumors for lung metastasis seeding through angiopoietin-like 4
.
Cell
2008
;
133
:
66
77
.
53.
Patel
SA
,
Nilsson
MB
,
Le
X
,
Cascone
T
,
Jain
RK
,
Heymach
JV
.
Molecular mechanisms and future implications of VEGF/VEGFR in cancer therapy
.
Clin Cancer Res
2023
;
29
:
30
9
.
54.
Zachary
I
,
Gliki
G
.
Signaling transduction mechanisms mediating biological actions of the vascular endothelial growth factor family
.
Cardiovasc Res
2001
;
49
:
568
81
.
55.
Bertucci
F
,
Ueno
NT
,
Finetti
P
,
Vermeulen
P
,
Lucci
A
,
Robertson
FM
, et al
.
Gene expression profiles of inflammatory breast cancer: correlation with response to neoadjuvant chemotherapy and metastasis-free survival
.
Ann Oncol
2014
;
25
:
358
65
.
56.
Ren
X
,
Zhang
L
,
Zhang
Y
,
Li
Z
,
Siemers
N
,
Zhang
Z
.
Insights gained from single-cell analysis of immune cells in the tumor microenvironment
.
Annu Rev Immunol
2021
;
39
:
583
609
.
57.
Hu
L
,
Su
L
,
Cheng
H
,
Mo
C
,
Ouyang
T
,
Li
J
, et al
.
Single-cell RNA sequencing reveals the cellular origin and evolution of breast cancer in BRCA1 mutation carriers
.
Cancer Res
2021
;
81
:
2600
11
.
58.
Slepicka
PF
,
Somasundara
AVH
,
Dos Santos
CO
.
The molecular basis of mammary gland development and epithelial differentiation
.
Semin Cell Dev Biol
2021
;
114
:
93
112
.
59.
Bambhroliya
A
,
Van Wyhe
RD
,
Kumar
S
,
Debeb
BG
,
Reddy
JP
,
Van Laere
S
, et al
.
Gene set analysis of post-lactational mammary gland involution gene signatures in inflammatory and triple-negative breast cancer
.
PLoS One
2018
;
13
:
e0192689
.
60.
Atkinson
RL
,
El-Zein
R
,
Valero
V
,
Lucci
A
,
Bevers
TB
,
Fouad
T
, et al
.
Epidemiological risk factors associated with inflammatory breast cancer subtypes
.
Cancer Causes Control
2016
;
27
:
359
66
.
61.
Asselin-Labat
ML
,
Vaillant
F
,
Sheridan
JM
,
Pal
B
,
Wu
D
,
Simpson
ER
, et al
.
Control of mammary stem cell function by steroid hormone signalling
.
Nature
2010
;
465
:
798
802
.
62.
Woodward
WA
.
Inflammatory breast cancer: unique biological and therapeutic considerations
.
Lancet Oncol
2015
;
16
:
e568
e76
.
63.
Yanagisawa
H
,
Komuta
Y
,
Kawano
H
,
Toyoda
M
,
Sango
K
.
Pleiotrophin induces neurite outgrowth and up-regulates growth-associated protein (GAP)-43 mRNA through the ALK/GSK3beta/beta-catenin signaling in developing mouse neurons
.
Neurosci Res
2010
;
66
:
111
6
.
64.
Sevillano
J
,
Sánchez-Alonso
MG
,
Zapatería
B
,
Calderón
M
,
Alcalá
M
,
Limones
M
, et al
.
Pleiotrophin deletion alters glucose homeostasis, energy metabolism and brown fat thermogenic function in mice
.
Diabetologia
2019
;
62
:
123
35
.
65.
Rosenfield
SM
,
Bowden
ET
,
Cohen-Missner
S
,
Gibby
KA
,
Ory
V
,
Henke
RT
, et al
.
Pleiotrophin (PTN) expression and function and in the mouse mammary gland and mammary epithelial cells
.
PLoS One
2012
;
7
:
e47876
.
66.
Christman
KL
,
Fang
Q
,
Kim
AJ
,
Sievers
RE
,
Fok
HH
,
Candia
AF
, et al
.
Pleiotrophin induces formation of functional neovasculature in vivo
.
Biochem Biophys Res Commun
2005
;
332
:
1146
52
.
67.
Wang
X
.
Pleiotrophin: Activity and mechanism
.
Adv Clin Chem
2020
;
98
:
51
89
.
68.
Chang
Y
,
Zuka
M
,
Perez-Pinera
P
,
Astudillo
A
,
Mortimer
J
,
Berenson
JR
, et al
.
Secretion of pleiotrophin stimulates breast cancer progression through remodeling of the tumor microenvironment
.
Proc Natl Acad Sci USA
2007
;
104
:
10888
93
.
69.
Ganguly
D
,
Schmidt
MO
,
Coleman
M
,
Ngo
TC
,
Sorrelle
N
,
Dominguez
ATA
, et al
.
Pleiotrophin drives a prometastatic immune niche in breast cancer
.
J Exp Med
2023
;
220
:
e20220610
.
70.
Blanco
R
,
Gerhardt
H
.
VEGF and Notch in tip and stalk cell selection
.
Cold Spring Harb Perspect Med
2013
;
3
:
a006569
.
71.
Van der Auwera
I
,
Van Laere
SJ
,
Van den Eynden
GG
,
Benoy
I
,
van Dam
P
,
Colpaert
CG
, et al
.
Increased angiogenesis and lymphangiogenesis in inflammatory versus noninflammatory breast cancer by real-time reverse transcriptase-PCR gene expression quantification
.
Clin Cancer Res
2004
;
10
:
7965
71
.
72.
Colpaert
CG
,
Vermeulen
PB
,
Benoy
I
,
Soubry
A
,
van Roy
F
,
van Beest
P
, et al
.
Inflammatory breast cancer shows angiogenesis with high endothelial proliferation rate and strong E-cadherin expression
.
Br J Cancer
2003
;
88
:
718
25
.
73.
Orekhov
AN
,
Bobryshev
YV
,
Chistiakov
DA
.
The complexity of cell composition of the intima of large arteries: focus on pericyte-like cells
.
Cardiovasc Res
2014
;
103
:
438
51
.
74.
Creazzo
TL
,
Godt
RE
,
Leatherbury
L
,
Conway
SJ
,
Kirby
ML
.
Role of cardiac neural crest cells in cardiovascular development
.
Annu Rev Physiol
1998
;
60
:
267
86
.
This open access article is distributed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0) license.