Endoglin (CD105) overexpression is associated with an immunosuppressive tumor microenvironment in murine lung cancer models and human lung adenocarcinoma Authors: Claudia Ollauri-Ibáñez, Blanca Ayuso-Íñigo, Inés Solano-SC, Paula Díez, Martín Perez-Andres, José M. Muñoz-Félix, Alicia Rodríguez-Barbero, Miguel Pericacho Published: 2026 ABSTRACT The tumor microenvironment (TME) is a major determinant of tumor response to different therapies, especially immunotherapy. Tumors with high CD8+ T-cell infiltration that respond well to immunotherapy are referred to as “hot” tumors, whereas “cold” tumors, with an immunosuppressive microenvironment, respond less effectively to this therapy. Endoglin (CD105) plays a critical role in angiogenesis, and its overexpression has been associated with a poorer prognosis in several types of cancer. This study aims to investigate whether high endoglin levels are associated with a cold microenvironment in tumors. Transgenic mice ubiquitously overexpressing human endoglin (ENG+) and wild-type C57BL/6J mice (WT) were used to analyze the TME in a Lewis Lung Carcinoma (LLC) subcutaneous xenograft model and in a lung cancer model. Tumors developed in ENG+ mice exhibited increased hypoxia, reduced CD8+ T-cell infiltration and an increased presence of immunosuppressive cells, such as M2 TAMs and Treg, compared to WT tumors. Thus, these tumors can be categorized as cold tumors. In addition, the analysis of the TME and endoglin expression in human lung adenocarcinoma samples showed that cold tumors have higher endoglin levels than hot tumors. These findings suggest that a hypoxic and immunosuppressive microenvironment may contribute to the poorer prognosis of tumors with high levels of this protein. This study highlights the potential of endoglin as a marker to predict the response to immunotherapy and to guide personalized treatment strategies in cancer patients.Supplementary InformationThe online version contains supplementary material available at 10.1038/s41598-026-55711-6. Introduction Antiangiogenic drugs became one of the great promises of cancer treatment when, in the 1970s, Folkman described neoangiogenesis as a critical process for the development of solid tumors1. Promising results from preclinical studies have led to their routine use in patients with various malignancies, including lung and renal cell cancer, among others. However, it has been demonstrated that they do not show the expected success in patients2,3. Instead, the lack of blood vessels creates a hypoxic environment that favors the proliferation of the most resistant and aggressive tumor cells4,5. In 2018, Dr. Alison and Dr. Honjo were awarded the Nobel Prize in Medicine for their research on immunotherapy6. This type of therapy harnesses and enhances the ability of immune system effector cells to selectively destroy malignant cells while preserving healthy tissue7. Immunotherapy is already a first-line treatment for non-small cell lung cancer8–11, as well as other solid tumors, but unfortunately, the objective response rate is only between 10% and 40%8,10. It has been shown that the response to these drugs depends largely on the nature of the tumor microenvironment (TME)12. The TME is formed by malignant and normal cells that constitute a tumor and their interactions through cytokines, chemokines, growth factors and enzymes that remodel the extracellular matrix13. Depending on the composition of the TME, there are two types of tumors: hot tumors, which respond well to immunotherapy, and cold tumors, which do not respond to this treatment. Hot or inflamed tumors are characterized by a TME with highly infiltrating anti-tumor cells, such as CD8+ T lymphocytes or M1 tumor-associated macrophages (TAMs), and low numbers of immunosuppressive cells, such as myeloid-derived suppressor cells (MDSCs), M2 TAMs or regulatory T lymphocytes (Tregs). In contrast, cold or non-inflamed tumors contain high numbers of tumor-promoting cells and low infiltration of anti-tumor immune cells12,14. Furthermore, these cold tumors are associated with poor vascular function, leading to poor perfusion and favoring the development of hypoxia, which promotes the survival of the most aggressive cells and the recruitment of anti-inflammatory cells4,12,15–17. For these reasons, identifying markers that can reveal the nature of the tumor and predict whether a patient will respond well or poorly to immunotherapy, from a simple biopsy, will prevent patients from undergoing ineffective treatments, avoiding possible side effects and bringing us closer to personalized medicine. One of these potential markers may be endoglin (CD105), a transforming growth factor-β (TGF-β) coreceptor that plays a critical role in various pathophysiological processes, including angiogenesis18,19, whose overexpression has been associated with poorer prognosis in different types of cancer20,21. Our group recently showed that endoglin overexpression preserves an active phenotype in endothelial cells, thereby preventing vascular maturation. In the case of tumors, this does not lead to increased vascularization or tumor mass growth, as expected, but rather to more vascular lacunae and hemorrhages. Moreover, mice overexpressing endoglin (ENG+) have more circulating tumor cells and develop more lung metastases, which may be responsible for the poorer prognosis of tumors with high endoglin levels22. Therefore, the aim of this work is to analyze whether elevated endoglin levels are associated with cold tumors. For this purpose, we will use the hEng transgenic murine model (ENG+) to investigate whether endoglin overexpression is correlated with more hypoxic tumors and an immunosuppressive TME, which, together with the alterations in angiogenesis already described, define tumors as cold. In addition, we will determine whether elevated endoglin expression is also associated with cold tumors in human lung adenocarcinoma samples. Materials and methods Mice All animal procedures were conducted in strict compliance with the European Community Council Directive (2010/63/EU) and Spanish legislation (RD1201/2005 and RD53/2013). The protocols were approved by the University of Salamanca Ethical Committee (Permit number #305). The study is reported in accordance with ARRIVE guidelines (https://arriveguidelines.org). Transgenic mice ubiquitously overexpressing endoglin (ENG+) were generated by the previous group leader via microinjection of a pCAGGS vector containing the complementary DNA sequence (cDNA) of human endoglin in the fertilized eggs of CBAxC57BL/6J mice, as previously described22,23 and are bred in the university animal facility. Wild-type C57BL/6J mice (WT) were used as controls. Animal selection was genotype-based, and no randomization or blinding was performed. Animals were housed under specific pathogen-free conditions at the University of Salamanca facilities in a temperature-controlled room with a 12-h light/dark cycle and reared on standard chow and water provided ad libitum. A similar number of 8–13-week-old male and female ENG+ and WT mice were used for the experiments. For animal anesthesia, 2% isoflurane in oxygen was used, and heat was provided during recovery. Animals were sacrificed via CO2 inhalation or cervical dislocation, depending on the experiment requirements. Human samples Access to human samples was granted by the Ethics Committee of the University of Salamanca (protocol code 437) within the legal framework of informed consent (Law 41/2002), the practice of biomedical research (Law 14/2007) and data processing and privacy in the national (Organic Law 3/2018) and European (2016/679/EU) context. Written informed consent was obtained from all donors for donation of samples to the biobank and their use in biomedical research. All methods involving human samples were performed in accordance with the Declaration of Helsinki and with all relevant guidelines and regulations. Fifty human samples of lung adenocarcinoma (LUAD) were provided by Red de Biobancos de Castilla y León (BEOCyL) as formalin-fixed paraffin-embedded (FFPE) blocks of tissue. All the samples were classified according to their histopathological pattern according to the WHO classification of 2021 by hematoxylin/eosin (H&E) staining. This classification was based on the analysis of 2 independent investigators under the supervision of a pathologist from the Hospital Universitario de Salamanca. Their vascular pattern was also analyzed and classified as angiogenic or non-angiogenic by immunohistochemistry with CD31 as an endothelial marker, and Weigert-Van Gieson staining to visualize the collagen and elastin fibers. The fifteen samples that were classified as solid adenocarcinomas and presented an angiogenic vascular pattern were selected for further analyses. Cell culture Lewis Lung Carcinoma (LLC) cells, provided by Dr. Aliño (Department of Pharmacology, University of Valencia, Spain), were cultured under adherent conditions in Dulbecco’s modified Eagle’s medium (DMEM) (Thermo Fisher Scientific) supplemented with 10% fetal bovine serum (FBS) (Thermo Fisher Scientific) and 50 U/mL penicillin–streptomycin (Thermo Fisher Scientific) and maintained in a 90% RH, 5% CO2 atmosphere at 37 °C. LLC cell subcutaneous xenograft model The subcutaneous xenograft model was performed as previously described22. Briefly, 106 LLC cells were resuspended in 50 µL of Matrigel® (Corning) and subcutaneously inoculated into the flanks of anesthetized mice. Two different tumors were induced in the dorsum of each mouse, with an initial diameter of 2–3 mm resulting from the cell–Matrigel® mixture. Ten days later, the mice were humanely euthanized via CO2 inhalation. The tumors were carefully removed and frozen in liquid nitrogen, fixed in 4% paraformaldehyde (PFA) or digested in collagenase, depending on the subsequent analysis. LLC cell lung model The lung tumor model was performed as previously described24. Briefly, 105 LLC cells were resuspended in 100 µL of PBS and injected intravenously into anesthetized mice. Previously, the mice were prewarmed for 10 min via a heating pad to induce vasodilatation. Fifteen days after tumor cell inoculation, the mice were humanely euthanized via cervical dislocation. Then, the trachea was cannulated, and the lungs were filled with 4% PFA to maintain tissue structure. The lungs were subsequently removed from the thoracic cavity and fixed in 4% PFA for 24 h. The lung lobes were dissected before being embedded in paraffin for histological analysis. Flow cytometry Half of the subcutaneous tumor was digested in 0.5 mg/mL collagenase from Clostridium histolyticum (Sigma‒Aldrich) and 0.02 mg/mL deoxyribonuclease I from bovine pancreas (Sigma‒Aldrich) for 1.45 h at 37 °C under agitation. The digestion mixture was then passed through a 70 μm-pore Falcon® cell strainer (Corning) and washed three times with PBS. The cell pellet was labeled with the antibody cocktail (Table S1) in the dark at room temperature for 15 min. After two washes with PBS, 106 cells were acquired on a BD FACSCanto™ II (BD Biosciences). Data were analyzed via Infinicyt 1.7 (Cytognos). Gating strategy and flow plots are shown in Figure S1 and the combination of markers used to identify each cell type is indicated in Table S2. Histological analysis: Immunofluorescence and immunohistochemistry Both subcutaneous and lung tumors destined for histological analysis were fixed in 4% PFA for 24 h. The samples were subsequently embedded in paraffin in the Comparative Molecular Pathology Unit of the Cancer Research Centre of Salamanca. Human samples were provided as formalin-fixed paraffin-embedded (FFPE) blocks of tissue. For immunohistochemistry, 2 μm sections were deparaffinized with xylene and rehydrated in a decreasing gradient of ethanol (100–70%). Antigens were unmasked with 10 mM sodium citrate buffer. The sections were incubated overnight at 4 °C with the primary antibody (Table S3) and for 1 h at room temperature with the secondary antibodies (Table S4). Following, they were incubated with the DAB Substrate Kit (Abcam ab64238) for up to 10 min. Finally, the nuclei were counterstained with hematoxylin, and the sections were dehydrated and mounted with DPX mounting medium (Casa Álvarez 10-8500). Images were acquired using an Olympus BX51 optical microscope and a PANNORAMIC Scan (3DHISTECH’s). For immunofluorescence, antigens from 2-µm sections were also unmasked with 10 mM sodium citrate buffer. The sections were incubated overnight at 4 °C with the primary antibody (Table S3) and in the dark for 1 h at room temperature with the secondary antibodies (Table S4). The cell nuclei were stained by incubating for 5 min with Hoechst before they were mounted with Prolong® Gold Antifade (Thermo Fisher Scientific). Images were acquired using an Axiovert 200 M fluorescence microscope (Nikon) and a Nanozoomer S60 (Hamamatsu). Image quantification Researchers who manually analyzed the images were blinded to the genotype of the samples. Subcutaneous xenograft model Ten random areas were selected at 20x magnification. Within each area, two subareas were marked around the blood vessels at a distance of three nuclei from the vessels (perivascular area). Then, within each subarea, vascular density and mural coverage were analyzed. For this purpose, the number of total blood vessels was quantified (based on the vessel marker Endomucin) with respect to the tumor area (in pixels), and the coverage of each vessel was also evaluated by assigning a score for the mural cell marker α-SMA (1 for completely mature vessels/high coverage, 0.5 for partially mature vessels, and 0 for immature vessels/low coverage). The final coverage percentage was then calculated as the ratio of the α-SMA score relative to the number of vessels in each area. Furthermore, to analyze perivascular lymphocyte infiltration, the number of CD3-positive cells was counted relative to the perivascular area (µm2). To analyze hypoxia, the areas positive for Glut1 or CaIX were selected with respect to the total tumor area (%). For cell infiltration analysis, several random areas were selected at 20x magnification, and the number of positive cells for CD8a, FoxP3, CD206 or Arg1 was counted. Lung model Tumors were fully photographed at 10x or 20x magnification, depending on the marker, and the entire tumor area was analyzed using Photoshop CS6 (in pixels). First, in each tumor, the angiogenic area and/or non-angiogenic area were selected. Tumors that presented at least 35% of the angiogenic area with respect to the total tumor area were considered angiogenic tumors. The remaining histological analyses were performed on the tumors identified as angiogenic, as in the subcutaneous model. Human lung adenocarcinoma Only samples previously classified as solid (growth pattern) and angiogenic (vascular pattern) were used in this study. The levels of endoglin expression were evaluated by assigning a microvascular density (MVD) score (0 for no staining, 1 for low expression and staining, 2 for medium expression, and 3 for high expression). For TME characterization, several random areas were selected, and the number of cells positive for CD8 or FoxP3 was counted. M2 TAMs infiltration was evaluated by measuring positive areas for CD206 with respect to the total tumor area (%). Tumors were classified as cold when CD8+ cell infiltration was low (CD8low) and hot when CD8high. For medium infiltration of CD8+ cells, Tregs and M2 TAMs were also considered: hot for low FoxP3 and/or CD206, and cold for high FoxP3 and/or CD206 (Fig. 6J). Cytokine gene expression (RT‒PCR) RNA was isolated from frozen tissue using NucleoSpin® RNA (Macherey-Nagel), and cDNA synthesis was subsequently performed with iScript RT Supermix (Bio-Rad) according to the manufacturer’s instructions. For qPCR, Supermix iQTM SYBR® Green (Bio-Rad) and an iQTM 5 System (Bio-Rad) were used. Gene expression results were normalized to ribosomal protein S13 (Rps13) and β-actin (Actb). The sequences or commercial references of the primers used are listed in Table S5. Statistical analysis All images shown are representative, and all the data are expressed as the means ± SEMs. The number of tumors included in each experiment is indicated in the corresponding figure legend. All graphical representations and statistical analyses were performed using Graph Pad 10.4.0 software. qPCR results are represented in box plots that show the median and the 25–75th percentiles, with whiskers showing the 10–90th percentiles. Histology quantification from human samples is represented in violin plots that show the median. Outliers were removed using the “Identify outliers” tool of the software. The Kolmogorov–Smirnov normality test was applied to the datasets before statistical comparisons. Unpaired t-test was used for Gaussian distributed datasets, whereas the Mann–Whitney test was used as a nonparametric test. Angiogenesis/co-option vascularization in lung tumors and hot/cold tumors in human samples were compared using Fisher’s exact test. In all cases, values were considered statistically significant when the p-value < 0.05 and are indicated with asterisks (*p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001). Results Continuous endoglin overexpression increases hypoxia within LLC subcutaneous tumors In this work, we first used a xenograft model in which LLC cells with Matrigel® are subcutaneously injected into the flank of ENG+ and WT mice (Fig. 1A). Immunohistochemical analysis revealed that only vessel-like structures, together with some infiltrating cells, are positive for human endoglin in the ENG+ mice (Fig. 1B). We analyzed vascular density and maturation by double immunofluorescence of the blood vessel marker endomucin and the mural cell marker alpha smooth muscle actin (αSMA)25. No differences in the number of endomucin-labeled blood vessels were found (Fig. 1C-D), but the average coverage of these vessels with αSMA-positive cells was significantly lower in tumors developed in ENG+ than in the WT mice (Fig. 1C and E). It is widely described that the lack of maturation of tumor blood vessels results in a dysfunctional and fenestrated vasculature that facilitates the intravasation of tumor cells, the generation of metastases and does not allow proper irrigation of tumors, leading to hypoxia26. To analyze the presence of hypoxic areas, we performed histological analyses of the hypoxic markers glucose transporter 1 (Glut1)27 and carbonic anhydrase IX (CaIX)5. We observed that the hypoxic area is larger in the subcutaneous tumors developed in ENG+ mice than in those in WT mice (Fig. 1F-H), confirming that incomplete maturation of blood vessels in ENG+ mice may affect the perfusion of the tumors and impair their correct oxygenation. Subcutaneous tumors developed in ENG+ mice show immunosuppressive tumor microenvironment Tumor hypoxia has been linked to a worse prognosis for several reasons, including that it can affect the recruitment and functionality of different immune cells to the TME. As a preliminary analysis of tumor immune cell infiltration, we analyzed the presence of different immune populations in subcutaneous tumors by flow cytometry of a single-cell suspension from the tumor. Before performing the tumor analysis, we confirmed, via peripheral blood analysis, that ENG+ mice did not show alterations in the analyzed populations (Figure S2). Following the analysis strategy outlined in Figure S1, we identified the most characteristic leukocyte populations: T lymphocytes (identifying CD4 and CD8), B lymphocytes, natural killer cells (NK), monocytes and neutrophils or MDSCs. Our results reveal a significant reduction in the number of NK cells in the tumors developed in ENG+ mice (Figure S3). Moreover, we also observed a clear trend toward less infiltration of CD8+ T cells, B cells and monocytes, but more infiltration of neutrophils or MDSCs (Figure S3). Fig. 1Endoglin overexpression decreases vessel maturation and increases hypoxia in tumors in a murine LLC subcutaneous tumor model. (A) Transgenic mice that overexpress endoglin (ENG+) and wild-type C57BL/6J mice (WT), as controls, were used to generate a subcutaneous xenograft tumor model with Lewis lung carcinoma cells (LLC). In this model, 106 cells in 50 µl of Matrigel® were subcutaneously inoculated in the flank of the mice, which were euthanized after 10 days. Created with BioRender.com. (B) Human endoglin immunostaining in tumor tissue. (C) Immunofluorescence of the vessel marker Endomucin (red) and the mural cell marker α-SMA (green). (D) Endoglin levels do not modify vascular density [n(WT) = 7; n(ENG+) = 9; p = 0,9419]. (E) Tumors from ENG+ mice have more immature vessels than WT mice [n(WT) = 7; n(ENG+) = 9; *p-value < 0,05]. (F) Immunofluorescence of the Glut1 marker and immunohistochemistry of the CaIX marker to study the hypoxic areas. (G) Tumors from ENG+ mice show a larger hypoxic area for the marker Glut1 than WT mice [n(WT) = 6; n(ENG+) = 9; *p-value < 0,05], ( H) also for the CaIX marker [n(WT) = 7; n(ENG+) = 9; *p-value < 0,05]. Endoglin overexpression decreases vessel maturation and increases hypoxia in tumors in a murine LLC subcutaneous tumor model. (A) Transgenic mice that overexpress endoglin (ENG+) and wild-type C57BL/6J mice (WT), as controls, were used to generate a subcutaneous xenograft tumor model with Lewis lung carcinoma cells (LLC). In this model, 106 cells in 50 µl of Matrigel® were subcutaneously inoculated in the flank of the mice, which were euthanized after 10 days. Created with BioRender.com. (B) Human endoglin immunostaining in tumor tissue. (C) Immunofluorescence of the vessel marker Endomucin (red) and the mural cell marker α-SMA (green). (D) Endoglin levels do not modify vascular density [n(WT) = 7; n(ENG+) = 9; p = 0,9419]. (E) Tumors from ENG+ mice have more immature vessels than WT mice [n(WT) = 7; n(ENG+) = 9; *p-value < 0,05]. (F) Immunofluorescence of the Glut1 marker and immunohistochemistry of the CaIX marker to study the hypoxic areas. (G) Tumors from ENG+ mice show a larger hypoxic area for the marker Glut1 than WT mice [n(WT) = 6; n(ENG+) = 9; *p-value < 0,05], ( H) also for the CaIX marker [n(WT) = 7; n(ENG+) = 9; *p-value < 0,05]. Once we verified that there were changes in the tumor immune populations, we studied the most relevant cell populations in the generation of the tumor immuno-microenvironment via immunohistochemistry. To analyze the presence of Tregs, we performed immunohistological staining of FoxP3, a characteristic marker of this cell type. We found a trend toward a greater number of these immunosuppressive cells in tumors from ENG+ mice than in those from WT mice, but the differences are not statistically significant (Fig. 2A-B). Moreover, to determine the infiltration of protumor M2 TAMs, we analyzed the expression of the specific markers CD206 and Arg1. In line with our hypothesis and the results already described the number of M2 TAMs in the subcutaneous tumors in ENG+ mice is greater than in WT mice (Fig. 2C-E). Fig. 2Endoglin overexpression favors protumoral components in the microenvironment of subcutaneous tumors. (A) FoxP3 immunostaining showing Treg lymphocytes infiltration. (B) Quantification of the number of Treg cells [n(WT) = 6 tumours, n(ENG+) = 8 tumours, p = 0,5277]. (C) M2 TAM infiltration observed by CD206 immunofluorescence (red) and Arg1 immunostaining. (D) Tumours from ENG+ mice show more infiltration of M2 TAM than tumours developed in WT mice, for both CD206 marker [n(WT) = 5, n(ENG+) = 9; *p-value < 0,05], and (E) Arg1 marker [n(WT) = 7, n(ENG+) = 10; *p-value < 0,05]. Endoglin overexpression favors protumoral components in the microenvironment of subcutaneous tumors. (A) FoxP3 immunostaining showing Treg lymphocytes infiltration. (B) Quantification of the number of Treg cells [n(WT) = 6 tumours, n(ENG+) = 8 tumours, p = 0,5277]. (C) M2 TAM infiltration observed by CD206 immunofluorescence (red) and Arg1 immunostaining. (D) Tumours from ENG+ mice show more infiltration of M2 TAM than tumours developed in WT mice, for both CD206 marker [n(WT) = 5, n(ENG+) = 9; *p-value < 0,05], and (E) Arg1 marker [n(WT) = 7, n(ENG+) = 10; *p-value < 0,05]. On the other hand, we also studied the presence of cells that inhibit tumor growth. We analyzed the presence of perivascular CD3+ lymphocytes. We observed that the tumors developed in ENG+ mice have fewer perivascular lymphocytes (Fig. 3A-B) and that these differences are markedly greater if we looked only at vessels with less than 25% mural coverage (Fig. 3A-D). Although CD3 is a pan-lymphocyte marker expressed by different subpopulations and maturational stages of lymphocytes, these perivascular cells are mainly cytotoxic lymphocytes28. To confirm that endoglin overexpression actually reduces the infiltration of these lymphocytes, which we previously observed via flow cytometry, we performed CD8 immunohistochemical staining. As expected, tumors developed in ENG+ mice have a lower number of cytotoxic lymphocytes than those developed in WT mice (Fig. 3E-F). Fig. 3Endoglin overexpression prevents infiltration of antitumoral components in subcutaneous tumors. (A) Immunofluorescence of the vessel marker Endomucin (red), mural cell marker α-SMA (yellow) and CD3+ lymphocytes (green). (B) Quantification of perivascular CD3+ T cells [n(WT) = 6 tumors, n(ENG+) = 8 tumors; *p-value < 0,05]. (C-D) Quantification of perivascular CD3+ T cells [n(WT) = 120 areas (from 6 tumors), n(ENG+) = 160 areas (from 8 tumors); *p-value < 0,05], (D) around vessels with less than 25% mural coverage [n(WT) = 34 areas (from 6 tumors), n(ENG+) = 76 areas (from 7 tumors); ****p-value < 0,0001], and (D) around vessels with 25–100% mural coverage [n(WT) = 86 areas (from 6 tumors), n(ENG+) = 84 areas (from 7 tumors); p = 0,5418]. (E) CD8a immunostaining showing cytotoxic lymphocyte infiltration. (F) Quantification of the number of CD8+ cells from infiltrated areas [n(WT) = 8 tumors, n(ENG+) = 10 tumors, *p < 0,05]. Endoglin overexpression prevents infiltration of antitumoral components in subcutaneous tumors. (A) Immunofluorescence of the vessel marker Endomucin (red), mural cell marker α-SMA (yellow) and CD3+ lymphocytes (green). (B) Quantification of perivascular CD3+ T cells [n(WT) = 6 tumors, n(ENG+) = 8 tumors; *p-value < 0,05]. (C-D) Quantification of perivascular CD3+ T cells [n(WT) = 120 areas (from 6 tumors), n(ENG+) = 160 areas (from 8 tumors); *p-value < 0,05], (D) around vessels with less than 25% mural coverage [n(WT) = 34 areas (from 6 tumors), n(ENG+) = 76 areas (from 7 tumors); ****p-value < 0,0001], and (D) around vessels with 25–100% mural coverage [n(WT) = 86 areas (from 6 tumors), n(ENG+) = 84 areas (from 7 tumors); p = 0,5418]. (E) CD8a immunostaining showing cytotoxic lymphocyte infiltration. (F) Quantification of the number of CD8+ cells from infiltrated areas [n(WT) = 8 tumors, n(ENG+) = 10 tumors, *p < 0,05]. Finally, we analyzed the expression of some antitumor cytokines by quantitative PCR, such as tumor necrosis factor (Tnf), interferon gamma (Ifng), interleukin 2 (Il2) and IL-12a (Il12a), and protumor cytokines, such as IL-1b (Il1b), IL-10 (Il10), IL-6 (Il6) and stromal cell-derived factor 1 (Cxcl12), and the expression of cyclooxygenase 2 (Cox2) (Figure S4). Taken together, except for Il12a, endoglin promotes the expression of protumoral cytokines, including Il6, Cxcl12 and Cox2 (Figure S4). Continuous endoglin overexpression increases hypoxia and shows immunosuppressive tumor microenvironment in lung tumors To test the effect of the overexpression of endoglin in tumors within the lung microenvironment, LLC cells were intravenously injected so that they travel through the bloodstream and implant in the lungs, where they generate tumors24(Fig. 4A). The lung is a highly vascularized tissue, where tumors are able to use at least two forms of vascularisation: vascular co-option, in which the tumor cells proliferate around existing blood vessels in the host tissue29, and angiogenesis. To differentiate between lung tumors with angiogenic vasculature and those with vascular co-option, double immunofluorescence of endomucin (blood vessel marker) and podoplanin (alveolar type I cell marker) was used. Co-opted areas, characterized by the maintenance of normal lung parenchyma with a “honeycomb” morphology30, present both markers inside the tumor mass; whereas angiogenic areas are characterized by the destruction of normal lung parenchyma and disorganized neovascularization within the tumor. Our results demonstrate that angiogenic, co-option and mixed tumors, which share both vascularization patterns, can be found in this model (Figure S5). Given that WT and ENG+ mice equally develop both angiogenic and co-option vascular profiles (Figure S5), in accordance with our objective, we focused only on angiogenic tumors. In this work, we considered an angiogenic tumor when the angiogenic area occupied at least 35% of the total tumor area. These angiogenic vessels present a similar structure to those observed in the subcutaneous xenograft model, so we first analyzed vascular density and vessel maturation by endomucin and α-SMA double immunofluorescence. As in the subcutaneous model, lung tumors developed in ENG+ and WT mice have a similar density of vasculature, with vessels from ENG+ mice having less mural coverage than those from WT mice (Fig. 4C-E). Accordingly, when we analyzed hypoxia, we detected that hypoxic areas are larger in ENG+ lung tumors than those developed in WT mice (Fig. 4F-H), confirming that endoglin overexpression also impairs correct oxygenation in this tumor model. Fig. 4Endoglin overexpression decreases vessel maturation and increases hypoxia in lung tumors. (A) Transgenic mice that overexpress endoglin (ENG+) and wild-type C57BL/6J mice (WT), as controls, were used to generate a murine lung tumor model with Lewis lung carcinoma cells (LLC). In this model, 105 cells resuspended in 100 µl of PBS were injected intravenously to travel through the bloodstream and implant in the lungs, where they generate the tumor. These mice were euthanized after 15 days. Created with BioRender.com. (B) Human endoglin immunostaining in the tumor tissue [scale bar: 50 μm]. (C) Immunofluorescence of the vessel marker Endomucin (red) and mural cell marker α-SMA (green). (D) Endoglin levels do not modify vascular density [n(WT) = 7; n(ENG+) = 10; p = 0,8868]. (E) Tumors from ENG+ mice have more immature vessels than those from WT mice [n(WT) = 10; n(ENG+) = 9; *p-value < 0,05]. (F) Immunofluorescence of the Glut1 marker and immunohistochemistry of the CaIX marker to study the hypoxic areas. (G) Tumors from ENG+ mice have a larger hypoxic area than those from WT mice [n(WT) = 22; n(ENG+) = 11; ***p-value < 0,0001], (H) also for the CaIX marker [n(WT) = 11 tumors; n(ENG+) = 7 tumors; ****p-value < 0,0001]. Endoglin overexpression decreases vessel maturation and increases hypoxia in lung tumors. (A) Transgenic mice that overexpress endoglin (ENG+) and wild-type C57BL/6J mice (WT), as controls, were used to generate a murine lung tumor model with Lewis lung carcinoma cells (LLC). In this model, 105 cells resuspended in 100 µl of PBS were injected intravenously to travel through the bloodstream and implant in the lungs, where they generate the tumor. These mice were euthanized after 15 days. Created with BioRender.com. (B) Human endoglin immunostaining in the tumor tissue [scale bar: 50 μm]. (C) Immunofluorescence of the vessel marker Endomucin (red) and mural cell marker α-SMA (green). (D) Endoglin levels do not modify vascular density [n(WT) = 7; n(ENG+) = 10; p = 0,8868]. (E) Tumors from ENG+ mice have more immature vessels than those from WT mice [n(WT) = 10; n(ENG+) = 9; *p-value < 0,05]. (F) Immunofluorescence of the Glut1 marker and immunohistochemistry of the CaIX marker to study the hypoxic areas. (G) Tumors from ENG+ mice have a larger hypoxic area than those from WT mice [n(WT) = 22; n(ENG+) = 11; ***p-value < 0,0001], (H) also for the CaIX marker [n(WT) = 11 tumors; n(ENG+) = 7 tumors; ****p-value < 0,0001]. We also analyzed the TME composition of these tumors via histological studies. Although lymphocyte infiltration is virtually undetectable, due to the short duration of tumor development and the lung tumor model used31, a trend toward less infiltration of antitumor CD8+ lymphocytes is observed in endoglin-overexpressing lung tumors (Fig. 5A-B). Conversely, and validating the results observed in the subcutaneous model, histological analyses of CD206 and Arg1 revealed that lung tumors that developed in ENG+ mice have a higher number of pro tumor M2 TAMs than WT tumors (Fig. 5C-E). Fig. 5Endoglin overexpression modifies the infiltration of leukocyte populations in angiogenic lung tumors. (A) CD8α immunostaining showing CD8+ lymphocyte infiltration. (B) Quantification of the number of CD8+ cells from infiltrated areas [n(WT) = 13, n(ENG+) = 10, p = 0,4176]. (C) CD206 immunofluorescence showing M2 TAM infiltration (red). (D) Tumors from ENG+ mice show greater infiltration of M2 TAMs than did those from WT mice [n(WT) = 5, n(ENG+) = 5; *p-value < 0,05], and also for (E) Arg1 marker [n(WT) = 12, n(ENG+) = 6; *p < 0,05]. Endoglin overexpression modifies the infiltration of leukocyte populations in angiogenic lung tumors. (A) CD8α immunostaining showing CD8+ lymphocyte infiltration. (B) Quantification of the number of CD8+ cells from infiltrated areas [n(WT) = 13, n(ENG+) = 10, p = 0,4176]. (C) CD206 immunofluorescence showing M2 TAM infiltration (red). (D) Tumors from ENG+ mice show greater infiltration of M2 TAMs than did those from WT mice [n(WT) = 5, n(ENG+) = 5; *p-value < 0,05], and also for (E) Arg1 marker [n(WT) = 12, n(ENG+) = 6; *p < 0,05]. High endoglin expression is associated with an immunosuppressive TME in solid angiogenic samples of human lung adenocarcinoma (LUAD) To analyze the clinical significance of endoglin expression in determining the TME, we study TME in lung adenocarcinomas (LUAD) previously classified as solid (according to the WHO 2021 classification of lung adenocarcinomas) and as angiogenic. Similar to what was done in mice, we used CD8, FoxP3 and CD206 as markers of the type of TME. The study of CD8+ cell infiltration shows that human tumors clustered into 3 subgroups, as shown in the violin plot (Fig. 6A). Some tumors had high infiltration (Fig. 6B), while others had very little infiltration (Fig. 6C), and a subgroup had intermediate infiltration. We therefore used the median to determine whether the values were high or low since, as explained above, high CD8+ lymphocyte infiltration is associated with a hot TME, whereas low infiltration is associated with a cold TME. To discriminate tumors with intermediate levels, we analyzed the infiltration of FoxP3- and CD206-positive cells associated with cold tumors. Similarly, we classified the samples as high or low, using the median in the violin plots (Fig. 6D and G), as we found tumors with high and low expression of these markers (Fig. 6E-F and H-I). Using the algorithm explained in the methods (Fig. 6J), we were able to classify 8 tumors as “hot” and 7 as “cold”. In one of the tumors, the result indicated by FoxP3 and CD206 was opposite to that of CD8, so we termed that tumor “unclassified” (Fig. 6K). We analyzed endoglin expression and assigned a score according to staining and MVD, and our results show that endoglin levels are higher in tumors classified as cold than in those classified as hot (Fig. 6L). On the other hand, when the endoglin score was represented in a violin plot, two highly defined populations were detected, allowing the tumors to be classified as Enghigh or Englow (Fig. 6M-O). Interestingly, lung adenocarcinomas classified as Enghigh correlate significantly more with those classified as cold, whereas Englow tumors correlate significantly more with those classified as hot (Fig. 6P). Discussion The recent, albeit unofficial, classification of tumors into two categories, “hot”, referred to as inflamed, and “cold”, referred to as immune-desert, is becoming increasingly widespread. This categorization depends mainly on the characteristics of its TME32,33: hot tumors are characterized by high infiltration of CD8+ T cells, as opposed to cold tumors, which are characterized by lower infiltration of CD8+ T cells and the presence of immunosuppressive cells such as MDSCs and Tregs. This composition of the TME affects tumor prognosis and response to immunotherapy, as cold tumors benefit less from this therapeutic approach and have a worse prognosis than hot tumors do, which usually respond well32–34. Cold tumors are also associated with poor vascular function, leading to poor perfusion and favoring the development of hypoxia, which promotes the survival of the most aggressive cells and the recruitment of anti-inflammatory cells4,12,15–17. Fig. 6Characterization of the tumor microenvironment in solid angiogenic samples of human lung adenocarcinoma. Characterization of the tumor microenvironment in solid angiogenic samples of human lung adenocarcinoma. (A) Distribution of the samples according to the number of CD8+ T cells per field. (B) CD8 immunostaining showing high and (C) low infiltration of CD8+ T cells [scale bar 20 μm]. (D) Distribution of the samples according to the number of Treg cells per field. (E) FoxP3 immunostaining showing high and (F) low infiltration of Treg cells [scale bar 50 μm]. (G) Distribution of the samples according to the area positive for CD206 with respect to the total tumor area. (H) CD206 immunostaining showing high and (I) low infiltration of M2 TAMs [scale bar 50 μm]. (J) Algorithm used for characterizing the type of TME: cold when CD8+ T-cell infiltration is low and hot when it is high, but for medium CD8+ T-cell, infiltration of Treg and M2 TAMs was also evaluated. Created with BioRender.com. (K) Classification of samples according to their type of TME (cold/hot/non-classified). (L) Tumors identified as cold correlate with higher levels of endoglin [n(cold) = 7, n(hot) = 8; *p < 0,05]. (M) Distribution of the different samples according to their MVD score. (N) CD105 immunostaining showing high and (O) low endoglin levels [scale bar 50 μm]. (P) Tumors with high endoglin levels are identified as cold, whereas tumors with low endoglin levels are hot according to their composition of the TME [n(cold) = 7, n(hot) = 8; *p < 0,05]. Characterization of the tumor microenvironment in solid angiogenic samples of human lung adenocarcinoma. Characterization of the tumor microenvironment in solid angiogenic samples of human lung adenocarcinoma. (A) Distribution of the samples according to the number of CD8+ T cells per field. (B) CD8 immunostaining showing high and (C) low infiltration of CD8+ T cells [scale bar 20 μm]. (D) Distribution of the samples according to the number of Treg cells per field. (E) FoxP3 immunostaining showing high and (F) low infiltration of Treg cells [scale bar 50 μm]. (G) Distribution of the samples according to the area positive for CD206 with respect to the total tumor area. (H) CD206 immunostaining showing high and (I) low infiltration of M2 TAMs [scale bar 50 μm]. (J) Algorithm used for characterizing the type of TME: cold when CD8+ T-cell infiltration is low and hot when it is high, but for medium CD8+ T-cell, infiltration of Treg and M2 TAMs was also evaluated. Created with BioRender.com. (K) Classification of samples according to their type of TME (cold/hot/non-classified). (L) Tumors identified as cold correlate with higher levels of endoglin [n(cold) = 7, n(hot) = 8; *p < 0,05]. (M) Distribution of the different samples according to their MVD score. (N) CD105 immunostaining showing high and (O) low endoglin levels [scale bar 50 μm]. (P) Tumors with high endoglin levels are identified as cold, whereas tumors with low endoglin levels are hot according to their composition of the TME [n(cold) = 7, n(hot) = 8; *p < 0,05]. In this study, we used two murine tumor models with LLC cells: a xenograft model implanted subcutaneously and a lung model after intravenous injection. Tumors created from LLC cells have been shown to have an immunosuppressive TME35,36. Our results show that endoglin overexpression increases this phenotype even more, as tumors developed in ENG+ mice present fewer anti-tumor cells, such as CD8+ lymphocytes and NK cells, and more pro-tumor Treg lymphocytes and M2 TAMs. The immunosuppressive phenotype is also reflected in TME cytokines, with increased expression of Cxcl12, Il-6 and Cox2, which promote tumor cell proliferation, migration, invasion and recruitment of M2 TAMs and Tregs37–39. The existence of an immunosuppressive TME in tumors with endoglin overexpression is consistent with the increased hypoxia observed in these tumors. This correlation has been highlighted when we have studied the TME and endoglin expression in human LUAD samples. As explained above, CD8+ cell infiltration is the main feature that allows tumors to be categorized as hot or cold12,32–34. For borderline phenotypes, we also used the presence of immunosuppressive cells, Tregs or M2 TAMs, as confirmation of the tumor phenotype12,32–34. Subsequent analysis of microvascular endoglin levels revealed that tumors classified as cold had higher endoglin levels than hot tumors. In addition, we observed that the samples were grouped as Enghigh or Englow and that Enghigh corresponded to cold tumors, whereas Englow corresponded to hot tumors. Several authors have shown that high microvessel endoglin levels are associated with a poor prognosis in cancers such as NSCLC40, hepatocellular carcinoma41, astrocytic tumors42 and breast carcinoma43, among others. Traditionally, it was assumed that the poorer prognosis of these tumors was due to increased angiogenesis leading to increased tumor growth. However, in a previous study, we demonstrated that endoglin overexpression does not increase angiogenesis but impedes proper capillary maturation. This impaired angiogenesis results in more permeable vessels that favor intravasation of tumor cells, generating metastases22. In this work, we demonstrate that tumors with elevated levels of endoglin are also associated with a hypoxic and immunosuppressive microenvironment, allowing us to classify these tumors as cold. The abundant literature on this subject allows us to propose that the alterations in angiogenesis caused by elevated endoglin expression22 are related to the appearance of hypoxic areas. Both alterations, especially hypoxia, prevent the recruitment of anti-tumor cells and favor the recruitment of pro-tumor cells, which defines the immunosuppressive microenvironment characteristic of cold tumors. In addition, hypoxia is able to induce endoglin expression, which could lead to a positive feedback loop44(Fig. 7). One limitation of this study is the use of a ubiquitous endoglin overexpression model rather than an endothelial-specific transgene, which prevents exclusive attribution of the effects to endothelial endoglin. Nevertheless, our prior in vitro and in vivo data show that endothelial cells overexpressing endoglin remain in an activated phenotype that impairs mural cell attachment22, the diminished CD3+ cell infiltration in the perivascular area of the immature vessel observed in this study, and the numerous clinical studies that localize endoglin predominantly to the tumor vasculature and associate higher vascular endoglin with poorer prognosis20,21 supports a primarily vascular contribution, while not excluding additional stromal inputs. In conclusion, our work proposes that endoglin could be used for both the diagnosis and treatment of cancer. In diagnosis, the use of endoglin as a biomarker could help to determine whether a tumor is cold or hot, making it possible to rule out cases in which immunotherapy would not be effective. In therapy, the use of anti-endoglin antibodies, such as carotuximab (TRC105), could help reduce endothelium activation, generating better blood vessels that could improve the delivery of other drugs to the tumor, decrease hypoxia, promote an immunostimulatory TME and reduce metastasis generation, resulting in a better response to immunotherapy and an overall better prognosis. Fig. 7Increased endothelial endoglin levels are related to altered angiogenesis, which results in hypoxia, that is associated with an immunosuppressive TME. Hypoxia also induces endoglin expression, which may also contribute to the sustained elevated endoglin levels. Increased endothelial endoglin levels are related to altered angiogenesis, which results in hypoxia, that is associated with an immunosuppressive TME. Hypoxia also induces endoglin expression, which may also contribute to the sustained elevated endoglin levels. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1