Crosstalk between tumor endothelial cells and cancer cells is important for metastasis initiation

Abstract

Background

Metastasis is a major contributor of cancer-related mortality and involves complex crosstalk between cancer cells and stromal cells modulated by various cytokines and growth factors. While the endothelium is essential for supplying nutrients and oxygen, little is known about its role in metastasis initiation. Determining the effect of endothelial-derived angiocrine factors on cancer cells may explain the mechanisms regulating metastasis initiation.

Results

In this study, we investigated the role of normal endothelial cells (NEC) and tumor endothelial cells (TEC) in regulating the rate-limiting steps of metastasis initiation. First, we demonstrated that TEC have a higher proliferation, migration, and angiogenic potential than NEC. TEC-conditioned media significantly promoted chemotaxis, invasion, and proliferation of cancer cells relative to NEC-conditioned media. Additionally, TEC facilitated faster cell-cell adhesion to tumor cells than NEC. Mass spectrometry analysis of endothelial cell secretome revealed higher levels of PDGF-AA, PDGF-C, and VEGFA in TEC-conditioned medium, which were associated with enriched PI3K-AKT, MAPK, and RAS signaling pathways, as well as regulation of actin cytoskeleton and focal adhesion. In vitro functional studies using recombinant proteins showed that PDGF-AA and PDGF-C significantly promoted cancer cell chemotaxis and invasion without affecting proliferation. However, unlike VEGFA, PDGF-AA and PDGF-C did not affect endothelial cell tube formation or vascular permeability. Interestingly, neutralization of TEC-derived PDGF-C signficantly inhibited tumor cell chemotaxis and invasion, and attenuated EphA2/AKT/P38/ERK signaling in vitro. In vivo, the co-injection of TEC and 4T1 cells resulted in significantly higher primary breast tumor growth and liver metastasis in an orthotopic mouse model.

Conclusion

Our results demonstrate that crosstalk between cancer cells and TEC, partly through angiocrine factors in the TME, induces the rate-limiting steps of metastasis initiation.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12964-025-02441-3.

Introduction

Metastasis is the main cause of cancer-related mortality, accounting for the majority of all cancer deaths [1–3]. Our ability to treat cancer effectively largely depends on our capacity to halt or eliminate metastasis. To develop effective diagnoses and therapies, it is essential to understand the molecular mechanisms within the tumor microenvironment (TME) that regulate metastasis initiation. Metastasis initiation involves several rate-limiting steps, including tumor cell invasion through the local extracellular matrix (ECM), chemotactic migration toward blood vessels, and adhesion to endothelium before intravasation and traveling to secondary organs [2–6]. The vascular endothelium plays a crucial role at both the primary and metastatic sites by facilitating angiogenesis, a phenomenon extensively regulated by a milieu of extracellular signals, including cancer cell-derived pro-angiogenic factors, cytokines, hypoxia, and ECM remodeling [5, 7, 8]. While the role of endothelial cells during angiogenesis has been well-characterized, emerging evidence suggests their function extends beyond the static vessel networks that deliver nutrients and oxygen.

Instead, endothelial cells assume a regulatory phenotype with privileged access to the tumor, orchestrating a unique cellular microenvironment by secreting tissue-specific paracrine factors, hereby termed ‘angiocrine factors’ that actively shape their local microenvironment and contribute to malignancy [9–12]. This is supported by multiple studies where direct co-culture of endothelial and cancer cells led to increased invasion, migration, and proliferation of tumor cells [10, 13–16]. However, whether the observed effects are due to angiocrine signaling or direct contact between tumor and endothelial cells is unclear. Furthermore, most co-culture studies involving endothelial cells focus on either their angiogenic role or tumor-endothelial crosstalk at metastatic sites [9, 11, 17] with limited information about the contribution of endothelial cells in metastasis initiation. Therefore, it is fundamental to study the direct effects of angiocrine factors on tumor cells and their contribution to metastasis initiation.

It is currently appreciated that the endothelium performs additional roles beyond its core angiogenic function. Endothelial cells are heterogeneous between tissues and organs, performing distinct physiological and pathophysiological functions [18–20]. Presumably, this heterogeneity is meant to serve the needs of each organ. However, most cancer research has relied on human umbilical vein endothelial cells (HUVECs) as in vitro cellular models of the endothelium and angiogenesis [21–23]. However, these cells reflect neither organo-specificity nor endothelial cell heterogeneity in the TME. There are now reported distinct biological differences between normal endothelial cells (NEC) and tumor endothelial cells (TEC), all of which reside in the TME and have distinct angiocrine signatures [19, 24–26]. TEC are considered irregular and abnormal and exhibit altered metabolic, secretomics, and transcriptomic profiles than their normal counterparts, based on the results of different single-cell transcriptomic studies [19, 24, 26–28]. These studies have laid the framework for comparative in vitro and in vivo studies of the different endothelial cells and their distinct functions in the TME, specifically the effects of their para-angiocrine signaling networks. Therefore, developing an understanding of the molecular mechanisms of endothelial cells in metastasis initiation beyond their role in malignant vascularization will provide an opportunity for the discovery of novel drug targets.

Here, we first provide an in-depth functional characterization of NEC and TEC and their subsequent regulation of the rate-limiting steps of metastasis initiation. Using in vitro functional assays and in vivo mouse models, we also describe their paracrine effect on tumor cells. Our results showed that TEC have a more proliferative, invasive, and angiogenic phenotype than NEC. Through their secretome, TEC actively drive an aggressive tumor phenotype by promoting the proliferation, chemotactic migration, invasion, and tumor-endothelial cell adhesion. We identified PDGF-AA and PDGF-C, two angiocrine factors highly secreted by TEC, as likely actors promoting tumor cell chemotaxis and invasion, important rate-limiting steps of metastasis initiation. In vivo, we found that TEC significantly promoted primary breast tumor growth and liver metastasis. These findings suggest that TEC establish a permissive microenvironment supporting metastasis initiation.

Materials and methods

Cell lines

The KPC cell line was isolated from a PDAC tumor and was a gift from the Tuveson Laboratory (Cold Spring Harbor Laboratory, United States) [29]. The 4T1 cell line was a kind gift from Fred Miller (Wayne State University). Both cells were cultured in DMEM GlutaMAX, supplemented with 10% FBS and 1% penicillin–streptomycin. The b.End5 cell line, herein referred to as the tumor endothelial cells (TEC), derived from brain endothelioma of BALB/c mouse, was a gift from the depositor (#96091930, Culture Collections UK) to the University of Copenhagen. BALB-5023, derived from BALB/c mouse primary brain microvascular endothelial cells, herein referred to as a normal endothelial cell (NEC), was purchased from Cell Biologics (Cat. BALB-5023). TEC was cultured in DMEM GlutaMAX with 10% FBS and 1% penicillin–streptomycin. NEC was cultured in an endothelial cell basal medium containing 5% FBS, VEGF, heparin, EGF, ECGS, hydrocortisone, and antibiotic–antimycotic (Cell Biologics, Cat. M1168). All cell Lines were cultured at 37 °C in a humidified atmosphere containing 5% CO₂ and routinely tested for mycoplasma.

Endothelial cell assays

Cell proliferation assay

Proliferation was determined using the CellTiter-Glo® Luminescent Cell Viability Assay (G7572, Promega) as per the manufacturer’s protocol. For endothelial cell proliferation, 1000 cells were plated in a 96-well plate in 100 µl of their respective Full growth media. Cells were allowed to settle overnight, and the 0-h time point was assessed at this time. Proliferation was monitored every 24 h by adding 100 µl CellTiter-Glo® reagent in each well, incubating for 10 min, and recording the luminescence in a plate reader. Proliferation was monitored for five days, and subsequent time points were normalized to 0-h values to assess proliferation.

Cell migration assays

Migration was investigated using the gap closure assay and the chemotactic transwell migration assay. A 2-well silicone insert (#81176, Ibidi) that produces a well-defined, reproducible gap was used for the gap closure assay. Before the gap closure assay, cells were starved for 18 h in 1% FBS growth medium. We seeded 5 × 10⁴ cells per insert and cultured overnight to allow them to settle. Afterward, the inserts were gently removed, and the cells were washed once with PBS and subcultured in their respective Full growth medium. The defined scratch areas were photographed at 0 and 16 h after removing the inserts. The area of the scratch was quantified using ImageJ software (NIH, Bethesda, MD, USA).

For chemotactic migration assay was conducted using the QCM Endothelial Cell Migration Kit (ECM200, Sigma Millipore) according to the manufacturer’s protocol. Before the assay, endothelial cells were starved for 18 h in 1% FBS growth medium. Afterward, they were harvested and subcultured in a fibronectin-coated transwell insert in serum-free media, with 10% FBS growth medium used as a chemoattractant. After 14 h, migrated cells were stained, washed twice in distilled water, and dried at room temperature for 10 min. Images of stained cells were captured using a light microscope. Tumor chemotaxis was subsequently evaluated by colorimetric analysis at OD 560 nm using a microplate reader.

Vascular permeability

Endothelial cell monolayer integrity was assessed using the in vitro vascular permeability assay kit (ECM644, Sigma Millipore) following the manufacturer’s instructions. Briefly, cells were cultured in their complete growth medium in a transwell insert for 48 h to form a continuous monolayer. Next, the endothelial monolayer was activated by two permeability factors: recombinant mouse TNF-α (#654245, Millipore) and VEGFA (#78102, Stem Cell Technologies) for 24 h. Respective serum-free culture media were used as controls. The inserts were washed twice with PBS, followed by the addition of 150 µl of FITC-Dextran solution (Diluted 1:40 in PBS) and incubated for 30 min at room temperature. Vascular permeability was evaluated by quantifying the amount of FITC-Dextran that passed through the endothelial monolayer by fluorescence (485 nm/535 nm). Vascular permeability of NEC and TEC was calculated relative to the permeability of the inactivated NEC monolayer.

Angiogenesis assays

The angiogenic properties of endothelial cells were assessed using the tube formation assay and the 3D spheroid sprouting assay. For the tube formation assay, 50 µl of chilled ECMatrix™ Solution (ECM625, Sigma Millipore) was added to multiple wells of a 96-well plate and incubated at 37 °C for 30 min to polymerize. Endothelial cells were trypsinized and counted, and 8000 cells in their respective serum-free medium were added to each well. Tube formation was assessed using 1 µg/ml recombinant mouse VEGFA used as a stimulant with respective serum-free medium as a control. Tube formation was evaluated by microscopy after 8 h, and the number of complete meshes, junctions, and tube length was quantified using ImageJ.

For spheroid sprouting assays, endothelial cell spheroids were generated using the hanging drop method. After 24 h, spheroids were harvested and embedded in a collagen matrix containing 1:1 20% FCS-methylcellulose and rat-tail type I collagen matrix (#3447-020-01, R&D Systems) in a low attachment 24-well plate. Spheroid sprouting was investigated in the presence of 1 µg/ml recombinant mouse VEGFA with respective serum-free medium as a control. After 24 h, sprouting was assessed microscopically, and the spheroid area was quantified using ImageJ.

Tumor cell assays

Preparation of endothelial cell conditioned medium

NEC and TEC conditioned medium were prepared by culturing 2 × 10⁶ cells in 10 cm dishes in their complete growth medium. At 80-90% confluence, endothelial cells were washed thrice with PBS and subcultured in 1% FBS + DMEM for 24 h. Endothelial cell conditioned medium was collected, filtered through 0.2 µm sterile filters, and used for subsequent functional assays. Fresh endothelial cell conditioned medium was used for all functional assays on tumor cells.

Cell proliferation

To determine the effect of endothelial cell secreted factors on cancer cell proliferation, KPC and 4T1 cells (1000 cells/well) were seeded in a 96-well plate in complete growth medium and allowed to settle overnight. Cell viability was assessed at this time point and assigned 0 h, and subsequent time points were normalized to 0-h values to assess proliferation. The complete growth medium was replaced by 100 µl fresh endothelial cell conditioned medium, and cell proliferation was assessed every 24 h (for five days) by adding 100 µl CellTiter-Glo® reagent in each well, incubating for 10 min, and recording the luminescence in a plate reader. Fresh endothelial cell conditioned medium was added every two days. 1% FBS-DMEM was used as a control.

Chemotaxis

KPC and 4T1 chemotactic migration were assessed using the colorimetric chemotaxis assay (ECM508, Millipore), according to the manufacturer’s instructions. Before the assay, tumor cells were starved for 18 h in serum-free medium. Afterward, starved tumor cells were harvested, counted, and 5 × 10⁴ cells were seeded in a transwell insert in serum-free medium. Fresh endothelial cell-conditioned medium was used as a chemoattractant, and 1% FBS-DMEM as a control. After 14 h, migrated cells were stained, washed twice in distilled water, and dried at room temperature for 10 min. Stained cells were imaged by alight microscope, and chemotaxis was subsequently quantified by colorimetric analysis at 560 nm using a microplate reader.

Tumor 3D invasion assay

KPC and 4T1 invasion were assessed using the spheroid invasion assay. As an initial step, tumor spheroids were generated using the hanging drop method. Spheroids were collected by washing in 10 ml PBS and centrifugation at 200 × g for 5 min. The supernatants were discarded, and spheroids were resuspended in 3 mg/ml rat-tail collagen I matrix and cultured in a low attachment 24-well plate. Fresh endothelial cell-conditioned medium was used as a stimulant, with 1% FBS-DMEM as a control. Spheroid invasion was assessed by microscopy after 24 h, and the invasion area was quantified using ImageJ.

Cell–cell adhesion assay

The adhesion of 4T1 and KPC cells to endothelial cell monolayers was assessed by microscopy and manual counting. TEC, KPC, and 4T1 reporter cells containing either GFP or RFP were generated using the CRISPR Knock-in system as previously described [30]. Full details are in the supplementary methods. We used the Cell Tracking Red Dye Kit (AB269446, Abcam) to label NEC because we were unable to tag them as was done for TEC, KPC, and 4T1 cells. For adhesion assays, 1 × 10⁵ labeled endothelial cells were seeded in a 6-well plate and cultured for 72 h to form a monolayer. GFP-labeled tumor cells were trypsinized, counted, and 1 × 10⁵ cells were seeded on the endothelial cell monolayer and incubated for one hour. Afterwards, wells were washed off three times using PBS to remove unattached cells. Cells were imaged (5 fields per group) using a fluorescent microscope, and tumor-endothelial cell adhesion was subsequently quantified by ImageJ.

Mass spectrometry secretome analysis

For sample preparation, NEC and TEC (1.5 × 10⁶ cells) were seeded in 10 cm cell culture dishes and cultured to 80% confluence. Afterwards, cells were washed twice with PBS to remove traces of FBS and then subcultured for 24 h in serum-free DMEM (Phenol red free). The medium was collected, filtered, and subjected to LC/MS analysis. Full details are in the supplementary methods.

Data analysis was conducted as previously described [31]. All raw LC-MS/MS data files were processed together using Proteome Discoverer version 2.4 (Thermo Fisher) with the use of label-free quantitation (LFQ) in both the processing and consensus steps. In the processing step, Oxidation (M), protein N-termini acetylation, and met-loss were set as dynamic modifications, with cysteine carbamidomethyl set as a static modification. All results were filtered with percolator using a 1% false discovery rate (FDR), and Minora Feature Detector was used for quantification. SequestHT was used as a database, matching spectra against the Mus musculus database from UniProt. Intensity normalization was performed in the total peptide amount mode. This involves summing the peptide group abundances for each sample, determining the maximum for all files, and then using it as a normalization factor. The dataset proteins were sorted for identification based on two or more unique peptides. To identify differentially expressed proteins, we used the LIMMA package (v 3.58.0) [32]. Proteins with detected intensities in at least two biological replicates were included in the analysis. Data was fit to the linear model, including the sample condition as a covariate. Differential gene usage was assessed with the eBayes function and adjusted for multiple comparisons with false discovery rate (FDR). Only proteins with a log2 FC of ± 1 and FDR < 0.05 were considered as significantly dysregulated. Further data analysis and visualization were done using GraphPad Prism (v 10.2.1), SR Plot [33] and R packages including ggplot2 (v 3.4.4) and heatmap3 (v 1.1.9). Protein-protein networks and functional enrichment analysis were performed using STRING (v 12.0).

Functional assays using recombinant PDGF-AA and PDGF-C

For all Functional assays, 100 ng/ml of either recombinant mouse PDGF-AA (SRP3228, Sigma) or mouse PDGF-C (#1447-PC-025, R&D Systems) was used to investigate their effect on tumor cell proliferation, invasion, and chemotaxis. In addition, the effect of recombinant PDGF-AA and PDGF-C on endothelial cell tube formation and vascular permeability was evaluated as previously described. Next, the signaling pathways activated by recombinant PDGF-AA and PDGF-C in KPC and 4T1 cells were investigated by western blotting. Full details are in the supplementary methods.

In vivo mouse models

All mouse experiments were conducted under authorization and guidelines approved by the Danish Inspectorate for Animal Experiments (License No. 2020-15-0201-00596). Aged-matched female Balb/c mice (6-12 weeks old) were used for orthotopic and experimental liver metastasis models. Orthotopic breast tumors were established by surgical implantation as previously described. [34]. Mice were randomly assigned to three groups (n = 5 for each group, repeated twice) and injected with 1 × 10⁵ 4T1 cells, a mixture of 1 × 10⁵ 4T1 cells and 2.5 × 10⁴ NEC or TEC (4:1) in 50 µl PBS in the left lower mammary fat pad. Tumor growth was monitored by palpation every 2 days for 28 days or a maximum volume of 1cm³. Tumor volume was calculated using the following formula: volume (mm³) = (length (mm) * width (mm)² × π)/6. The disease burden was assessed by measuring the weight of mice during the experiment. At the study endpoint, tumors were resected and weighed, and lungs and livers were collected and processed for immunohistochemistry analysis. Some mice (n = 2 for 4T1, n = 1 for 4T1 + NEC, and n = 2 for 4T1 + TEC) were excluded from subsequent data analysis as they had no visible tumors at the study endpoint. An additional mouse (n = 1 for 4T1) was also eliminated from subsequent data analysis by outlier analysis (ROUT method, GraphPad Prism). The final data analysis used n = 7 for 4T1, n = 9 for 4T1 + NEC, and n = 8 for 4T1 + TEC. Liver and lung metastatic burden was quantified by counting and averaging the number of all metastatic foci in three non-consecutive H&E liver or lung sections.

Experimental liver metastasis was done using intrasplenic injection as previously described [35]. Briefly, mice were randomly assigned to three groups (n = 5 per group, repeated twice). Briefly, 5 × 10⁵ 4T1 cells, a 4:1 mixture of 5 × 10⁵ 4T1 cells and 1 × 10⁵ NEC or TEC in PBS, were injected into the spleen of immunocompetent Balb/c mice using a Hamilton 29-G syringe. Mice were monitored by weighing every 4 days up to 21 days. At the study endpoint, mice were euthanized, and livers and lungs were collected, weighed, and processed for immunohistochemical analysis. Metastatic burden was assessed by quantifying the size of all metastatic foci (in three non-consecutive H&E liver sections) as a percentage of the entire liver section. Lung metastatic burden was assessed by counting and averaging the number of all metastatic foci in three non-consecutive H&E lung sections.

Statistical analysis

GraphPad Prism version 10.4.1(627) was used for statistical analysis. All in vitro data were performed in biological triplicates, and error bars on bar plots indicate mean + SD unless stated otherwise. For all in vitro proliferation and in vivo tumor growth assays, growth differences were estimated by calculating the area under the curve (AUC) method, followed by significance analysis. For analyses of two groups, an unpaired Student’s t-test with Welch's correction was used. For comparisons involving more than two groups, one-way ANOVA paired with Tukey’s multiple comparison test was used. A p-value of less than 0.05 was considered significant, and p-values were denoted as follows: *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Results

TEC have an aggressive phenotype compared to NEC

In this study, we evaluated the differences between normal and tumor endothelial cells. We used the well-characterized cell line b.End5 as a tumor endothelial cell line and herein referred to as TEC. BALB-5023 cells are derived from brain vasculature and considered normal endothelial cells, and are referred to as NEC for simplicity. We characterized NEC and TEC (Supplementary Fig. 1A) by evaluating their proliferation, migration, vascular integrity, and angiogenic potential. By immunoblotting, we showed that both cells express known endothelial-specific markers such as CD31, VCAM1, and Endomucin (Fig. 1A). Using qPCR, we also showed that both NEC and TEC express Vegfr1, Vegfr2, and Vegfr3 (Supplementary Fig. 1B). These receptors mediate VEGF signaling, which is vital for physiological and pathological vascularization. Next, we assessed the proliferation differences of these cells using the luminescence assay for five days. While all the cells had a similar growth profile in the first 72 h, afterwards, TEC had a significantly higher proliferation than NEC (Fig. 1B). We then investigated the motility differences of these endothelial cells using chemotactic migration and gap closure assays. While there was no difference in their chemotactic migration capacity (Supplementary Fig. 1C), TEC migrated significantly faster compared to NEC (Fig. 1C-D). This difference in motility between TEC and NEC is consistent with previous reports that showed glioblastoma-derived ECs having a higher motility than their normal counterparts [36].

Fig. 1.

Functional characterization of endothelial cells. A Expression of different endothelial cell markers in NEC and TEC. β-Actin is the loading control for VCAM1 and the sample control for both CD31 and Endomucin. B Graph showing endothelial cell proliferation. Growth curve was calculated using the area under the curve (AUC) method. C Images of the gap closure assay. D Quantified area of gap closure for NEC and TEC after 16 h. E Quantification of the vascular permeability of endothelial monolayers after treatment with recombinant TNF-α or VEGFA. Vascular permeability was plotted relative to the value of the inactivated NEC monolayer. F Images of endothelial cell tube formation. G Quantification of the total tube length in NEC and TEC treated with control or recombinant VEGFA for 8 h (n = 8 images per independent experiment, 3 biological repeats). H Images of NEC and TEC spheroids sprouting post-treatment with recombinant VEGFA for 24 h. I Quantification of the sprouting area of NEC and TEC spheroids seen in H (n = 8 images per independent experiment, 3 biological repeats). All data are expressed as mean ± SD of three independent experiments, and statistical significance is determined using the unpaired Student’s t-test: *p < 0.05, **p < 0.01, and ****p < 0.0001

As vascular permeability is indispensable for tumor metastasis, we evaluated endothelial cell permeability using the FITC-Dextran transwell assay. We reasoned that tumor cells secrete factors that communicate with the endothelial barrier through paracrine signaling to facilitate extravasation and metastatic dissemination. There was no difference in the baseline permeability of both cell lines without activation. However, when either endothelial cell monolayer was activated using recombinant TNF-α or VEGFA, we reported an increase in the permeability of TEC compared to NEC (Fig. 1E). This higher permeability of TEC post-activation points towards a metastatic dissemination role.

The development and sprouting of new capillaries from resident endothelial cells are the hallmarks of angiogenesis [20, 25]. We evaluated the angiogenic potential of these endothelial cells using the tube formation and the spheroid sprouting assays. Even in the absence of an angiogenic stimulant, TEC showed a significantly higher tube formation potential than NEC, a phenotype enhanced by the addition of recombinant VEGFA (Fig. 1F). We reported a significant increase in tube length as well as the number of meshes and junctions in TEC than NEC (Fig. 1G, Supplementary Fig. 1E-F). Similarly, TEC had a significantly higher sprouting capacity than NEC, which was also amplified in the presence of recombinant VEGFA (Fig. 1H-I).

Given these differences between NEC and TEC, we were curious whether activation of NEC using tumor-conditioned medium would alter their phenotype and assume TEC characteristics. We investigated changes in chemotaxis and vascular permeability of activated endothelial cells, as there were no differences between NEC and TEC at baseline. Additionally, we also characterized the proliferation of activated endothelial cells relative to the inactivated ones. Our findings showed that activated NEC had a significantly higher chemotactic migration capacity and was more permeable than inactivated NEC. Nevertheless, the chemotactic and permeability capacity of activated NEC were still lower than those of activated TEC (Supplementary Fig. 1G-H). Additionally, we compared the short-term proliferation of activated and inactivated EC. We observed that activated EC proliferated faster than their inactivated counterparts. Interestingly, activated TECs were still highly proliferative compared to activated NEC (Supplementary Fig. 1I).

TEC-conditioned medium promotes cancer cell proliferation, chemotaxis, and invasion

To understand the role of angiocrine factors on KPC and 4T1 cells, we used NEC- and TEC-conditioned media to investigate tumor cell proliferation, chemotaxis, and invasion in vitro. First, we evaluated the effect of endothelial cell conditioned media on tumor cell proliferation. Conditioned medium from both endothelial cell lines stimulated tumor proliferation compared to the control. However, TEC-conditioned medium induced significantly more proliferation than NEC-conditioned medium in both cancer cells (Fig. 2A, B). Next, we evaluated the effect of endothelial conditioned medium on the chemotaxis and invasion of KPC and 4T1 cells. Chemotaxis of tumor cells in the TME and invasion through local ECM are critical steps during metastatic initiation [1, 4]. Both endothelial cell conditioned media stimulated the chemotactic migration of KPC and 4T1 cells compared to the control, with TEC-conditioned medium having the greatest chemotaxis effect (Fig. 2C-D). Similarly, TEC-conditioned medium significantly upregulated the invasion capacity of 4T1 and KPC cells compared to NEC-conditioned medium (Fig. 2E, F). These findings are in line with a recent study that showed conditioned medium from glioblastoma-derived endothelial cells promoted the proliferation and migration of patient-derived gliomaspheres in vitro [37].

Fig. 2.

TEC-conditioned medium promotes chemotaxis and invasion. A, B Graphs showing the proliferation of KPC and 4T1 cells cultured using endothelial cell conditioned medium for 5 days. Growth curve differences were calculated using the area under the curve (AUC) method. C Images of chemotactic cancer cells. D Graph showing the chemotactic migration of KPC and 4T1 cells in the presence of endothelial cell conditioned medium quantified by colorimetric analysis. E Images of 3D tumor spheroids invasion after 24 h following treatment with endothelial cell conditioned medium. F Quantification of the area invaded by tumor spheroids (n = 8 images per independent experiment, 3 biological repeats). G Immunofluorescence images of cell-cell adhesion with white arrows depicting tumor cells adhering to endothelial cells. H Quantification of the number of adhered tumor cells per field (white arrows in G, n = 5 fields per independent experiment, 3 biological repeats). All data are expressed as mean ± SD of three independent experiments, and statistical significance is determined using the unpaired Student's t-test: *p < 0.05, **p < 0.01, and ****p < 0.0001

Therefore, the adhesion of tumor cells to the endothelium is a critical rate-limiting step in metastasis initiation. We investigated the adhesion of KPC and 4T1 cells to both endothelial cell monolayers. Briefly, GFP-labeled cancer cells were added to a monolayer of growing RFP-labeled endothelial cells and cultured for an hour. Our results showed that TEC monolayers significantly promoted cell-cell adhesion of tumor cells compared to NEC monolayers (Fig. 2G-H). Notably, KPC and 4T1 cells displayed a similar adhesion profile to each endothelial cell line. Taken together, our results demonstrate that TEC-conditioned medium contains factors that significantly promote the aggressiveness of KPC and 4T1 cells.

Secretome profiling of NEC and TEC

We conducted a secretome analysis of NEC and TEC conditioned media using label-free quantitative proteomics to gain insight into the angiocrine factors and molecular mechanisms via which endothelial cells influence tumor cell properties. We identified more than 2300 proteins across each of the three replicates per endothelial cell line (Fig. 3A, Supplementary Table 1), with 2280 proteins commonly identified in both endothelial cell lines (Supplementary Fig. 2B-C). By hierarchical clustering and principal component analysis, we showed that NEC and TEC samples cluster separately, confirming their differential secretome composition despite a common organ of origin (Fig. 3B, Supplementary Fig. 2C).

Fig. 3.

Secretome profiling of NEC and TEC. A Bar graph showing the number of proteins quantified in each sample. B Heatmap comparison showing the secretome profiles of NEC and TEC. C Bar graph showing the number of differentially extracellular proteins in TEC-conditioned medium compared to NEC-conditioned medium. 62 proteins were highly secreted in TEC while 92 proteins were highly secreted in NEC. D Pie chart of the distribution of matrisome proteins as part of the all the proteins identified in the extracellular space. E Volcano plot showing the altered secreted proteins in NEC- and TEC-conditioned medium. Data points in blue and red show the proteins highly secreted in NEC- and TEC-conditioned medium, respectively. F Heatmap showing expression of the 32 secreted proteins identified in NEC- and TEC-conditioned medium (Red-highly secreted in TEC, Blue-highly secreted by NEC, Grey-no differences). G Chord diagram showing the enriched KEGG pathways associated with only the differentially secreted factors (9 in TEC-conditioned medium and 10 in NEC-conditioned medium) shown in Fig. 3E-F

To determine the differentially secreted proteins in the extracellular space, we focused on common proteins identified in both endothelial cell secretomes. As an initial step, we characterized the subcellular location of the commonly identified proteins using Uniprot ID mapping. Approximately 318 proteins (14%) of all commonly identified proteins were in the extracellular space (Supplementary Fig. 2D). Using a cut-off of p < 0.05 and FC ± 2.0, we identified 154 differentially regulated proteins (62 upregulated and 92 downregulated) in TEC vs NEC (Fig. 3C, Supplementary Table 2). Next, we analyzed enriched molecular functions and KEGG pathways associated with proteins significantly secreted by NEC and TEC. For TEC, among the relevant and significantly enriched molecular functions were PDGF binding, lysyl oxidase activity, chemoattractant activity, growth factor/receptor activity, ECM structural constituents, proteoglycans, laminin, and collagen binding. For NEC, the relevant enriched molecular functions include chemorepellent activity, endopeptidase activity, endopeptidase inhibitor activity, and the binding of IGF receptors, semaphorin receptors, and integrins (Supplementary Fig. 2E, Supplementary Table 3). For KEGG pathways, we found an enrichment of multiple cancer-associated pathways in TEC, including ECM-receptor interaction, focal adhesion, amoebiasis, proteoglycans in cancer, resistance to EGFR inhibitors, and different signaling pathways, such as PI3K-AKT, MAPK, and Relaxin pathways (Supplementary Fig. 2 F, Supplementary Table 3). The most prominent KEGG pathway for NEC was the complement and coagulation cascade, alongside some cancer-associated pathways such as PI3K-AKT signaling, focal adhesion, and ECM-receptor interaction.

Some macromolecules comprising the secretome include ECM molecules and secreted factors that collectively make up the matrisome [38]. A previous report by Naba and colleagues predicted the in silico matrisome by characterizing the composition of ECM molecules and associated factors in humans and mice, which is currently referred to as the NABA matrisome [39]. The NABA matrisome is divided into two major groups, namely the core matrisome consisting of collagens, proteoglycans, and ECM glycoproteins, and matrisome-associated proteins composed of secreted factors, ECM regulators, and ECM-affiliated proteins [38, 39]. As NEC and TEC have a common biological origin, we compared their secretome to characterize their matrisome differences. We focused on the 318 proteins in the extracellular space (Supplementary Fig. 2D) and found that 172 proteins were members of the NABA matrisome (Fig. 3D, Supplementary Table 4-Sheet 1). Of these, 44.8% were core matrisome proteins, including collagen (20), proteoglycans (7), and ECM glycoproteins (50), while 55.2% were matrisome-affiliated proteins, including ECM regulators (46), ECM-affiliated proteins (17), and secreted factors (32). However, of the 154 differentially regulated proteins identified in our study, 96 were members of the Naba matrisome, with 58 being non-NABA members (Supplementary Table 4-Sheet 2).

Secreted factors are the largest component of the NABA matrisome and constitute growth factors, cytokines, metabolites, chemokines, and hormones [38, 39]. Crosstalk between tumor cells and stromal cells in the TME is mediated via some of these secreted factors, which promote aggressive malignancy. Our results showed 32 secreted factors, based on the Naba matrisome nomenclature, with nine members highly secreted by TEC and 10 members overly secreted by NEC (Fig. 3E-F). An in-depth analysis of the 32 secreted proteins revealed the presence of platelet-derived growth factor (PDGF) family members in the endothelial secretome. PDGF-C and PDGF-AA were significantly secreted by TEC, PDGF-D was significantly secreted by NEC, with no difference in PDGF-BB secretion (Supplementary Table 5). Protein-protein interaction and functional enrichment analyses of the differentially secreted proteins showed an enrichment of different cancer-associated signaling networks. Focal adhesion, chemokine signaling, regulation of actin cytoskeleton, Rap1 signaling, and pathways in cancer were uniquely enriched in TEC with PDGF-AA, PDGF-C, and VEGFA as the dominant factors involved. In NEC, TGF-β signaling was the only uniquely enriched signaling pathway (Fig. 3G, Supplementary Table 6). In both NEC and TEC, MAPK signaling, PI3K-AKT signaling, resistance to EGFR inhibitors, RAS signaling, and cytokine-cytokine receptor interaction were examples of commonly enriched pathways. We concluded that these differences in NEC and TEC secretomes are partly responsible for their inherent phenotypic differences and the observed functional effect they exert on cancer cells.

PDGF-AA and PDGF-C promote tumor chemotaxis and invasion

For further downstream analysis, we selected factors significantly elevated in TEC, as conditioned medium from this endothelial cell line induced significant biological differences in tumor cells in vitro. We focused on PDGF-C and PDGF-AA as they belong to the same family, and both were highly secreted by TEC. First, we assessed the expression of PDGFRA, the principal receptor for PDGF-AA and PDGF-C, in tumor cells. Our immunoblot results showed that KPC cells did not express PDGFRA but expressed EphA2 and NRP1 (Fig. 4A), receptors that have been previously been shown to interact with PDGF-AA and PDGF-C, respectively [40, 41]. Using ELISA, we quantified the levels of PDGF-AA and PDGF-C in the conditioned medium of NEC and TEC. Our results showed that TEC significantly secreted more PDGF-AA and PDGF-C than NEC (Fig. 4B, C), validating our mass spectrometry findings.

Fig. 4.

PDGF-AA and PDGF-C promote chemotaxis and invasion. A Immunoblot with indicated antibodies in lysates from KPC and 4T1 cells. Secreted levels of PDGF-AA (B) and PDGF-C (C) in NEC and TEC evaluated using ELISA. D, E Cell proliferation of KPC and 4T1 stimulated with control, recombinant PDGF-AA or PDGF-C. F Images of KPC and 4T1 cells chemotaxis in the presence of control, recombinant PDGF-AA or PDGF-C as chemoattractants. (G) Quantification of tumor cell chemotaxis. H Images of KPC and 4T1 spheroid invasion after 24 h of treatment with control, recombinant PDGF-AA or PDGF-C. I Quantification of the area invaded by tumor spheroids (n = 8 spheres per independent experiment, 3 biological repeats). All data are expressed as mean ± SD of three independent experiments (except Fig. 4B where n = 2), and statistical significance is determined using the unpaired Student's t-test: *p < 0.05, **p < 0.01, and ***p < 0.001

First, we investigated the effect of these recombinant proteins on cell proliferation. In both cell lines, neither PDGF-AA nor PDGF-C promoted cell proliferation (Fig. 4D-E). As chemotaxis and invasion are essential steps in metastasis initiation, we investigated the effect of recombinant PDGF-AA and PDGF-C on these steps in vitro. We showed that recombinant PDGF-AA and PDGF-C significantly promoted tumor cell chemotaxis (Fig. 4F-G) and invasion (Fig. 4H-I). Additionally, we characterized the autocrine role of recombinant PDGF-AA and PDGF-C on endothelial cells. Neither recombinant protein affected the vascular permeability of either NEC or TEC (Supplementary Fig. 3A), suggesting that other secreted factors like VEGFA play this role as described in Fig. 1E. Moreover, as sustained angiogenesis is a hallmark of cancer and vital in metastasis initiation [4, 6, 12], we evaluated whether recombinant PDGF-AA and PDGF-C could stimulate TEC tube formation. We showed that these recombinant proteins did not promote TEC tube formation (Supplementary Fig. 3B-C). These results suggest PDGF-AA and PDGF-C, significantly abundant in TEC-conditioned medium, act in a paracrine manner and are partly responsible for increased tumor cell chemotaxis and invasion.

We evaluated the possible intracellular signaling pathways induced by recombinant PDGF-AA and PDGF-C based on their effect in promoting tumor cell invasion and chemotaxis. Based on our secretome analysis, AKT and MAPK pathways were some of the enriched pathways regulated by these two factors. Treatment of KPC and 4T1 cells with these recombinant proteins increased the levels of active pP38 (T180/Y182), pAKT (S473), and pERK1/2 (T202/Y204). Interestingly, we also observed an increased pEphA2 (S897) in 4T1 cells when treated with both recombinant proteins (Fig. 5A). Next, we investigated whether inhibiting EphA2 and PDGFRA receptors with ALW-II-41-27 (An Eph family tyrosine kinase inhibitor, which has been demonstrated to block EphA2 signaling) and Cediranib (a VEGFR and PDGFR inhibitor), would block downstream EphA2/P38/AKT/ERK signaling. As an initial step, we evaluated whether these drugs would inhibit tumor chemotaxis. While ALW-II-41-27 inhibited chemotaxis in both cancer cells, Cediranib only inhibited 4T1 chemotaxis with no effect in KPC cells (Fig. 5B-C). Inhibition of 4T1 cells by ALW-II-41-27, an EphA2 inhibitor, was associated with decreased activation of pEphA2 (S897), pP38 (T180/Y182), pAKT (S473), and pERK1/2 (T202/Y204) phosphorylation in the presence of either PDGF-AA or PDGF-C (Fig. 5D-E, Supplementary Fig. 3D-E). While Cediranib treatment had a similar inhibitory effect in 4T1 cells, it generally did not affect PDGF-C mediated pP38 (T180/Y182) phosphorylation (Supplementary Fig. 3E). For KPC cells, a consistent observation across the biological replicates conducted was the decreased PDGF-activated pP38 (T180/Y182) phosphorylation in the presence of ALW-II-41-27, as well as a decrease in PDGF-C mediated pP38 (T180/Y182) activation in the presence of Cediranib (Supplementary Fig. 3F-G). This may partly explain the decreased chemotaxis of KPC cells in the presence of this inhibitor. Surprisingly, there was also a consistent increase in phosphorylation of pEphA2 and pERK1/2 in KPC cells after treatment with ALW-II-41-27, in the presence of either PDGF ligand (Supplementary Fig. 3F-G).

Fig. 5.

Inhibition of paracrine PDGF and VEGF signaling in cancer and endothelial cells. A Immunoblot with indicated antibodies in lysates from KPC and 4T1 cells. β-Actin is the loading control for AKT and P38 and the sample control for EphA2 and ERK1/2. Full immunoblots are shown in the supplementary information file. B Images of tumor chemotaxis. Chemotaxis was stimulated using recombinant PDGF-AA or PDGF-C in the presence of DMSO, 1 µM ALW-II-41-27 or 1 µM Cediranib and evaluated after 14 h. C Quantification of tumor chemotaxis. D Immunoblot with indicated antibodies in lysates from 4T1 cells stimulated with recombinant PDGF-AA in the presence of DMSO, 1 µM ALW-II-41-27 or 1 µM Cediranib. β-Actin is the loading control for AKT and P38 and the sample control for EphA2 and ERK1/2. E Immunoblot with indicated antibodies in lysates from 4T1 cells stimulated with recombinant PDGF-C in the presence of DMSO, 1 µM ALW-II-41-27 or 1 µM Cediranib. β-Actin is the loading control for EphA2 and P38 and the sample control for AKT and ERK1/2. F Quantification of the tumor-conditioned medium mediated TEC permeability in the presence of 10 nM Cediranib, with 1% FBS-DMEM as control. Vascular permeability was plotted relative to value of untreated TEC monolayer. G Graph showing tumor-conditioned medium mediated TEC proliferation in the presence of 10 nM Cediranib, with 1% FBS-DMEM as control. Proliferation differences were calculated using area under the curve (AUC) method. H Images of TEC tube formation. I Quantification of TEC tube length treated with KPC- or 4T1-conditioned medium, with serum-free medium as control, in the presence DMSO or 10 nM Cediranib after 8 h (n = 6 per independent experiment, 3 biological repeats). All data are expressed as mean ± SD of three independent experiments, and statistical significance is determined using the unpaired Student's t-test: *p < 0.05, **p < 0.01, and ***p < 0.001

As Cediranib inhibits VEGFRs as well as PDGFRs, we investigated its effect on tumor vascularization in the presence of tumor-conditioned medium. We reasoned that tumor-conditioned medium contains high levels of VEGFA and blocking VEGFA-VEGFR signaling would inhibit endothelial functions. ELISA analysis showed that tumor cells secreted more VEGFA than endothelial cells (Supplementary Fig. 3H), and that pre-treatment of either KPC or 4T1 cells with recombinant PDGFs did not elevate VEGFA secretion (Supplementary Fig. 3I). While tumor-conditioned medium promoted endothelial permeability and proliferation, treatment with 10 nM Cediranib significantly attenuated both processes (Fig. 5F-G). Both KPC and 4T1 conditioned medium significantly increased TEC tube formation, which was partially inhibited by 10 nM Cediranib treatment (Fig. 5H-I).

Neutralization of PDGF-C inhibits tumor chemotaxis and invasion

To further validate the effect of PDGFs secreted by TEC, we investigated the effect of TEC-conditioned medium on tumor cells in the presence of neutralizing antibodies against PDGF-AA and PDGF-C. As an initial step, we investigated the effect of TEC-derived PDGF-AA or PDGF-C on tumor chemotaxis and invasion in the presence of their respective neutralizing antibodies. PDGF-C neutralization significantly inhibited tumor chemotaxis and invasion in both KPC and 4T1 cells (Fig. 6A-D). On the contrary, PDGF-AA neutralization marginally but significantly reduced chemotaxis and invasion in 4T1 cells without affecting KPC cells. Next, we evaluated the effect of PDGF neutralization on intracellular signaling using immunoblotting. In both cell lines, neutralization of TEC-derived PDGF-C inhibited EphA2 (S897) phosphorylation, which subsequently led to decreased AKT/P38/ERK signaling compared to the IgG control (Fig. 6E-F). Interestingly, while PDGF-AA neutralization did not affect these signaling pathways in KPC cells, there was a consistent marginal decrease in the levels of pEphA2 (S897) and pP38 (T180/Y182) in 4T1 cells (Supplementary Fig. 3J-K). These findings suggest that TEC-derived PDGF-C is a key mediator of chemotaxis and invasion in this setting.

Fig. 6.

Neutralization of TEC-derived PDGF-C inhibits tumor chemotaxis and invasion. A Images of KPC and 4T1 chemotaxis. Chemotaxis was stimulated using TEC-conditioned medium pre-incubated with neutralizing antibodies against PDGF-AA and PDGF-C and evaluated after 14 h. B Quantification of tumor cell chemotaxis. C Images of KPC and 4T1 spheroid invasion after 24 h of treatment using TEC-conditioned medium pre-incubated with neutralizing antibodies against PDGF-AA and PDGF-C. D Quantification of the area invaded by tumor spheroids (n = 8 per independent experiment, 3 biological repeats). E Immunoblot with indicated antibodies in lysates from KPC cells stimulated with TEC-conditioned medium pre-incubated with neutralizing antibodies against PDGF-AA or PDGF-C. β-Actin is the loading control for P38 and the sample control for EphA2, AKT and ERK1/2. F Immunoblot with indicated antibodies in lysates from 4T1 cells stimulated with TEC-conditioned medium pre-incubated with neutralizing antibodies against PDGF-AA or PDGF-C. β-Actin is the loading control for P38 and EphA2 and the sample control for AKT and ERK1/2. All data are expressed as mean ± SD of three independent experiments, and statistical significance is determined using the unpaired Student's t-test: *p < 0.05, **p < 0.01, and ***p < 0.001

TEC promote primary tumor growth and spontaneous liver metastasis in an orthotopic breast model

We evaluated the in vivo effects of the co-injection of 4T1 and endothelial cells on tumor growth and metastasis using an orthotopic and liver metastasis models. In the orthotopic model, no significant differences were reported in endpoint tumor weight, endpoint tumor volume, as well as liver/lung/spleen weight between the three groups (Supplementary Fig. 4A-C). However, four weeks after orthotopic injection, we observed striking differences in the tumor growth between the 4T1, 4T1 + NEC, and 4T1 + TEC mice. Notably, we found that mice injected with 4T1 + TEC had a significantly higher tumor growth rate than mice injected with either 4T1 + NEC or 4T1 cells only (Fig. 7A). This higher growth rate of 4T1 + TEC tumors was also associated with an increased KI67 immunostaining (Fig. 7B-C). One of the objectives of this study was to determine whether the co-injection of endothelial cells with 4T1 would modulate the angiogenic potential of the developing tumors in vivo. While there was no significant difference in the number of CD31+ vessels between the mice groups co-injected with endothelial cells, 4T1 + TEC tumors had significantly more vessels than 4T1 only tumors (Fig. 7B, D). Further, immunohistochemical analysis of orthotopic tumors showed significantly increased expression of PDGF-AA and PDGF-C in the 4T1 + TEC group compared to the 4T1 group only (Fig. 7E-F). While there was no significant difference in the expression levels of PDGF-AA and PDGF-C between the mice co-injected with either endothelial cell line, 4T1 + TEC tumors had higher levels of PDGF-AA and PDGF-C than 4T1 + NEC tumors (Fig. 7E-F). In addition, 4T1 + TEC tumors had significantly higher expression levels of pEphA2 (S897), pAKT (S473), and pP38 (T180/Y182) compared to 4T1 + NEC or 4T1 only tumors. (Fig. 7E-F). However, there was no difference in the expression of pERK1/2 (T202/Y204) across these three groups.

Fig. 7.

TEC promote primary tumor growth and spontaneous liver metastasis in an orthotopic breast model. A Tumor growth curve of mice with orthotopic breast tumors co-injected with either NEC or TEC. Data is expressed as mean ± SEM and tumor growth differences were determined using area under curve (AUC) method. B Immunohistochemical staining of KI67 and CD31 in tissue sections of resected orthotopic tumors. C Quantification of KI67+ (proliferation), and D CD31+ (vessels) in tissue sections of resected orthotopic tumors. E Immunohistochemical staining of PDGF-AA, PDGF-C, pEphA2, pAKT, pP38, and pERK1/2 in tissue sections of resected orthotopic tumors. F Quantification of the staining intensities in the tissue sections of orthotopic tumors of the targets shown in E. Quantification was done by averaging the intensities of the three different regions per resected tumor section. GH&E images of liver sections of orthotopic mice from the three experimental groups with arrows showing metastasis lesions in the liver. H Quantification of the number of metastasis lesions in the livers of orthotopic mice shown in G. Quantification was done by averaging the number of metastatic lesions in three non-consecutive H&E liver sections per mouse. All other data is expressed as mean ± SD, and statistical significance was determined using one-way ANOVA paired with Tukey’s multiple comparison test: *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001

Aggressive breast cancer is associated with liver metastasis, a phenotype commonly recapitulated in orthotopic mouse models. To this end, we evaluated the liver metastatic burden between the three groups. 4T1 + TEC mice had a higher liver metastasis burden than 4T1 + NEC and 4T1 mice based on the average number of metastatic lesions counted in three non-consecutive H&E liver sections (Fig. 7G-H). However, this was not the case for experimental liver metastasis, as all groups had similar areas of liver metastasis foci post-intrasplenic injection (Supplementary Fig. 4D-E). These findings demonstrate that TEC at the primary tumor site are essential mediators of liver metastasis. Further, we evaluated the lungs of mice at endpoint for both experimental models, as the lung is also a common site of breast cancer metastasis. For the orthotopic model, there were no differences between the three groups with regard to the number of metastatic foci in the lungs (Supplementary Fig. 4F-G). In the intrasplenic injection model, no differences in the lung metastasis burden were observed between the 4T1 + NEC and the 4T1 + TEC groups. Surprisingly, the co-injection of 4T1 + NEC led to a significantly higher number of metastatic lung foci compared to the 4T1 group. Nevertheless, on average, 4T1 + TEC mice had more metastatic foci than the 4T1 group, albeit not statistically significant (Supplementary Fig. 4H-I).

Discussion

Much remains to be elucidated about the determinants of metastasis initiation. While most studies have focused on tumor cell autonomous pathways, emerging evidence shows that stromal cells, including endothelial cells, play a significant role in initiating and sustaining the metastatic process. This partly explains why anti-angiogenic drugs have limited clinical efficacy, with a subset of patients developing outright resistance while others acquire it during treatment [11, 15, 42]. Endothelial cells are important components of the TME that not only play a role in angiogenesis but also orchestrate and regulate their surrounding microenvironment through angiocrine factors. While many elements of the endothelial-tumor cell crosstalk have been identified, their role in metastasis initiation is an underexplored area that will lay a foundation for future translational studies.

In the current study, we provide a detailed characterization of the NEC and TEC and their role in regulating metastasis initiation. Historically, there have been assumptions that endothelial cells have a similar molecular profile to vascular networks that feed the tumor as they arise from the surrounding normal tissue. However, our findings show that TEC have a distinct phenotype from their counterpart NEC in terms of proliferation, migration, angiogenic capacity, and vascular leakiness. These results align with two previous studies that compared NEC, TEC from high-metastatic tumors, and TEC from low-metastatic tumors [27, 43]. TEC from high-metastatic tumors were highly proliferative, invasive, and more responsive to VEGFA than either NEC or TEC from low-metastatic tumors [27]. Our study had similar observations, with TEC being more responsive to VEGFA than NEC, and significantly promoting angiogenesis and vascular permeability. Endothelium-tumor crosstalk via angiocrine factors at the primary site likely plays a significant role in metastasis initiation. Previous studies have demonstrated the paracrine role of the angiocrine factors in facilitating metastasis [10, 13, 44]. Similarly, we showed that TEC conditioned medium significantly promoted tumor cell proliferation, chemotaxis, and invasion compared to NEC conditioned medium. Tumor cell adhesion to the endothelium is a critical step that facilitates intravasation and migration to secondary sites [2, 6, 7]. We showed that TEC facilitated faster tumor cell adhesion than NEC, corroborating findings from a previous study [43]. Moreover, factors such as CX3CL1, CXCL12, VEGFA, and S100A10, which we identified in our research as highly secreted by TEC, have also previously been shown to promote the adhesion of tumor cells to the endothelium [45–48]. As tumor cell chemotaxis, invasion, adhesion to endothelium, and vascular permeability are rate-limiting steps of metastasis initiation, our findings support the concept that angiocrine factors secreted by TEC significantly regulate these initial steps of metastasis. Even though our study does not use breast- or pancreas-specific endothelial cells, these findings justify the need for further validation using organ-specific endothelial cells.

To understand the molecular mechanisms mediating the observed biological effects of angiocrine factors on tumor cells, we carried out a secretome analysis of NEC- and TEC-conditioned medium. The factors expressed and secreted into the conditioned medium by the respective endothelial cells could explain their phenotypic differences and their paracrine effect on tumor cells. Our secretome analysis showed that PDGF-AA, PDGF-C, and VEGFA were among the angiocrine factors significantly secreted by TEC and were associated with enriched cancer signaling pathways such as focal adhesion, PI3K-AKT, and MAPK signaling. Importantly, the results of our secretome analysis point towards TEC heterogeneity and possibly an organo-specific angiocrine signature. An earlier study identified biglycan as a highly secreted molecule by TEC, which promotes metastasis in a subcutaneous mouse model [43]. Yet, in this study, we found biglycan to be highly secreted by NEC, a finding that underscores TEC heterogeneity and organo-specific angiocrine signature. These differences warrant further characterization of organ-specific TEC to find either unique or common markers that can be targeted therapeutically.

PDGF-AA and PDGF-C are members of the PDGF family frequently expressed in different tumors and correlate with tumor invasiveness, drug resistance, and poor clinical outcomes [49–51]. As growth factors, they have multifunctional roles in both normal physiology and cancer, and in the latter, they target stromal, vascular, and tumor cells to modulate tumor metastasis. Functionally, while recombinant PDGF-AA and PDGFC were not essential for tumor proliferation, angiogenesis, and vascular permeability in our study, both factors significantly promoted tumor chemotaxis and invasion, two critical rate-limiting steps of metastasis initiation. These findings strengthen their paracrine role on tumor cells, and are in line with previous studies that have implicated both factors as regulators of ECM invasion and chemotaxis [51, 52]. Mechanistically, we showed that recombinant PDGF-AA and PDGF-C activated pP38, pAKT, and pERK1/2 in both tumor cell lines, key players in PI3K/AKT/MAPK signaling pathways that are likely to regulate invasion and chemotaxis. Further, both recombinant proteins also activated pEphA2 signaling in 4T1 cells only, suggesting the heterogeneity and complexity of cellular signaling. Moreover, neutralization of TEC-derived PDGF-C significantly inhibited chemotaxis and invasion in both cell lines and was associated with decreased phosphorylation of these proteins. Interestingly, while PDGF-AA neutralization did not affect KPC invasion and chemotaxis, it significantly, but to a lesser extent than PDGF-C neutralization, attenuated 4T1 chemotaxis and invasion. This is partly because of the higher levels of PDGF-C than PDGF-AA in TEC-conditioned medium, which likely overrides the effect of PDGF-AA neutralization. Importantly, these mechanistic results validate our secretome data, which associated high PDGF secretion with enriched PI3K-AKT and MAPK signaling. Therefore, high secretion of PDGF-AA and PDGF-C by TEC into the TME partly explains the aggressive and tumor-promoting phenotype associated with TEC in this study [43].

To underscore the significance of PDGF-AA and PDGF-C signaling in this context, we inhibited their functions by using ALW-II-41-27 (An EphA2 inhibitor) and Cediranib (A VEGFR and PDGFR inhibitor). In 4T1 cells, both drugs reduced PDGF-AA- and PDGF-C-mediated chemotaxis, which was also associated with decreased activation of pEphA2, pAKT, and pERK1/2 signaling. This activation of EphA2 in 4T1 cells by either PDGF protein corroborates previous studies that demonstrated EphA2 receptor promiscuity independent of its primary ligand Ephrin-A1 [40, 53, 54]. Interestingly, while both inhibitors reduced pP38 levels in the 4T1 cell line post PDGF-AA treatment, Cediranib inhibition did not affect pP38 levels in the presence of PDGF-C. We speculate that PDGF-C signaling in 4T1 cells might be via PDGFRA and independent of P38, which might explain why it remains unchanged upon PDGFRA inhibition. These observations in 4T1 cells were validated using neutralizing antibodies, where TEC-derived PDGF-C neutralization inactivated intracellular signaling and attenuated tumor chemotaxis and invasion.

While TEC-derived PDGF-C neutralization also inhibited EphA2/AKT/P38/ERK signaling in KPC cells, the phosphorylation levels of these proteins remained unaffected, decreased, or marginally increased in the presence of inhibitors. While we observed a PDGF-dependent increase in pAKT, pP38, and pERK1/2 levels, treatment with ALW-II-41-27, an EphA2 inhibitor, in the presence of either PDGF ligand led to a consistent increase in pEphA2 and pERK1/2 levels. Interestingly, Cediranib treatment also elevated pEphA2 levels in the presence of either PDGF ligand in KPC cells. These cell-dependent observations in KPC cells are complex but not surprising. Accumulating evidence shows that pEphA2 (S897) activation is important for therapy resistance in various solid tumors, and is mediated by downstream AKT/MAPK activation [55]. In melanoma cells, pEphA2 (S897) activation was associated with increased pAKT signaling in a ligand-independent manner, following BRAF inhibition [56]. Similarly, pEphA2 (S987) activation promoted the growth and metastasis of sunitinib-resistant renal carcinoma cells in vivo, and this was associated with increased activation of ERK/AKT/STAT3 signaling [57]. Moreover, pEphA2 (S897) activation has been reported as a response to therapy treatment in nasopharyngeal, liver, and ovarian cancers, which matches our findings [58–60]. Based on this evidence, pEphA2 (S897) activation and the subsequent increase in pAKT/pERK levels in KPC cells may be a response to the drug treatment. However, these complex observations require further in-depth studies to validate our speculations.

The tumor angiogenic switch is an important step that activates local vasculature to facilitate metastasis. Our in vitro data show that, unlike VEGFA, PDGF-AA and PDGF-C are not involved in angiogenesis as they neither promote vascular leakiness nor enhance endothelial tube formation in vitro. As TEC secrete more VEGFA into the TME than NEC, it augments what is already secreted by tumor cells to enhance pathological angiogenesis and vascular leakiness. Blocking tumor cell-derived VEGFA signaling using Cediranib decreased endothelial proliferation, tube formation, and vascular permeability, key processes that promote tumor vascularization. Taken together, we hypothesize that in a growing primary tumor, pathological angiogenesis, tumor chemotaxis, and invasion are in part mediated by VEGFA, PDGF-AA, and PDGF-C cooperative signaling, leading to metastasis initiation.

It has been demonstrated that a permissive microenvironment regulated by stromal cells such as CAFs, endothelial, and immune cells likely leads to a high metastatic burden [10, 11, 17, 61]. Using an in vivo model, we demonstrated that orthotopic 4T1 + TEC tumors had a higher growth rate than either 4T1 + NEC or 4T1 only tumors and a significantly higher metastatic burden in the liver. However, this was not the case in an experimental liver metastasis model, suggesting that cellular crosstalk at the primary site is a crucial regulator of metastasis initiation. Notably, in the orthotopic mouse model, there were no differences in lung metastasis burden in part because of the short experimental duration. Given these initial findings, in the future, we propose to orthotopically implant 4T1 xenografts, surgically remove the tumor after 28 days, and monitor the mice further to allow for significant metastasis development. This will also allow us to validate in-depth the contribution of NEC and TEC in metastasis development at distant sites. Regardless, the higher growth rate of 4T1 + TEC tumors was also associated with the activation of EphA2/AKT/P38 signaling, as demonstrated by their increased phosphorylation status in the extracted tumors. Intriguingly, this significantly high activation of EphA2/AKT/P38 in 4T1 + TEC tumors may be in part due to the high expression of PDGF-AA and PDGF-C in these tumors, as demonstrated by immunohistochemical staining. These results corroborate our in vitro experiments, where we also demonstrated PDGF-mediated activation of EphA2/AKT/P38 signaling. Additionally, 4T1 + TEC tumors had significantly more blood vessels than control tumors. These observations validated our in vitro findings that demonstrated TEC's aggressive phenotype, which they confer on tumor cells via angiocrine factors. In addition, these findings are in line with other published studies that have reported the metastasis-promoting effect of TEC in an in vivo setting [27, 43, 62]. Additional studies are necessary to define and understand crosstalk between TEC and cancer cells in vivo and whether repressing or activating different angiocrine factors at the primary site have an effect on distant metastasis.

Nevertheless, our study has certain limitations. First, the endothelial cells used in the study are derived from the brain and do not represent the organ-specificity of our tumor cells. Given the distinct nature of the brain TME compared to other organs, it is highly likely that the angiocrine factors secreted by either breast- or pancreas-derived EC are different from brain-derived EC. As this was an exploratory study, in the future, we will use organ-specific endothelial cells to validate these findings. Second, the intricacies around PDGF-AA/PDGF-C signaling in KPC cells, coupled with the ineffectiveness of ALW-II-41-27 and Cediranib, warrant further investigation. Third, our in vivo studies hypothesize that the co-injected endothelial cells can secrete similar factors as they do in vitro and subsequently instruct the microenvironment to activate metastasis. While we do observe differences between our experimental groups, we do not know how the orthotopic breast microenvironment and resident endothelial cells affect our cells post-injection, and the contribution of direct cell–cell contact or factor-mediated paracrine signaling to these findings. Moreover, as we did not use labeled cells, it is difficult to distinguish how NEC and TEC integrate with the breast vascular system and how this promotes disease progression. Conversely, the complexity of these findings requires further in-depth investigations to address the current study’s limitations.

In summary, our findings provide insights into the mechanisms of cellular crosstalk between cancer cells and TEC that are likely to regulate metastasis initiation. We demonstrate that TEC secrete more PDGF-AA and PDGF-C, which promote tumor chemotaxis and invasion, and VEGFA, which may augment what is already secreted by tumor cells to drive pathological angiogenesis and vascular leakiness. Targeting pathways activated by these factors attenuated chemotaxis and angiogenesis, demonstrating that angiocrine factors instruct their niche environments and should be explored as drug targets. Efforts are ongoing in the laboratory to confirm these findings in more complex in vivo models to elucidate the significance of angiocrine signaling in metastasis initiation.

Authors' contributions

VOO- conceived the project, planned the research, conducted or participated in all experiments, analyzed the data, and wrote the paper. JLC, JPN, AGM- performed mouse experiments and reviewed the manuscript, FIJ- conducted in vitro experiments, MVL- carried out the mass spectrometry data acquisition, LB- bioinformatic analyses of the proteomics data, MKGH- performed all the IHC stainings, KJM- conducted in vitro experiments and reviewed the manuscript, JTE- funding acquisition, conceived the project, supervised the work, and writing of the manuscript. All authors read, reviewed, and approved the final version of the manuscript.

Funding

Open access funding provided by Copenhagen University. VOO and JTE are funded by the European Union’s Horizon 2020 Research and Innovation Programme under the Marie Skłodowska-Curie grant agreement no. 945322 and the Danish Cancer Society Grant No. R302-A17400. JLC is funded by the European Union’s Horizon 2020 research and innovation program under the Marie Sklodowska-Curie grant agreement number 101034291 (DISCOVER).

Data availability

All data generated during this study are included in the article or supplementary figures. All data from the corresponding authors is available upon request. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD062200 (Reviewer username: reviewer_pxd062200@ebi.ac.uk, Password: v0G1asgj6bFv).

Declarations

Ethics approval and consent to participate

All mouse experiments were conducted under authorization and guidelines approved by the Danish Inspectorate for Animal Experiments (License No. 2020-15-0201-00596).

Consent for publication

Not applicable.

Competing interest

The authors declare no competing interests.