Diabetic hyperglycemia-induced glycation regulates tumor vasculature integrity via NF-κB-mediated GM-CSF secretion by tumor cells

Summary

Glycation is recognized for its ability to stiffen collagen matrices through non-enzymatic crosslinking. However, the biological effects of glycation in vivo, particularly in diabetic hyperglycemia remain less understood. Clinical observations suggest that diabetes contributes to pathogenic tumor vasculature, yet the underlying mechanisms are not clear. Here, we employed in vitro hydrogels, an ex ovo chicken chorioallantoic membrane model, a mouse model, and RNA sequencing to demonstrate how diabetic hyperglycemia impacts tumor vasculature barrier integrity via glycation-mediated ECM stiffening and production of advanced glycation end products. Stiffened ECM and activated AGE-RAGE signaling lead to heightened contractility, activation of NF-κB, increased secretion of GM-CSF in tumor cells, and subsequently, impaired tumor vasculature barrier function. Our findings reveal a novel mechanism by which matrix stiffness and AGE-RAGE signaling synergistically govern vasculature barrier integrity and highlight a key role for matrix stiffness in NF-κB activity.

Graphical Abstract

eTOC blurb

Wang et al. described the mechanism by which diabetic hyperglycemia impairs tumor vasculature integrity via glycation. They discovered that the glycation-induced stiffened ECM and activated AGE-RAGE signaling lead to heightened cell contractility, activation of NF-κB, and increased secretion of GM-CSF in tumor cells, resulting in impaired tumor vascular barrier function. They emphasized the crucial role of glycation in the interaction between diabetes and breast cancer and proposed the broader potential impact of glycation in disease progression goes beyond breast cancer.

Introduction

Glycation is a non-enzymatic reaction where sugars react with proteins, lipids, or nucleic acids¹. This process is particularly valued in the field of tissue engineering for its ability to modify the mechanical properties of collagen matrices. Through glycation, collagen matrices can be stiffened, a modification crucial for numerous biomedical applications, without significantly altering other mechanical properties such as elasticity or porosity^2,3. Despite widespread use and effectiveness in biomaterials, glycation in vivo remains incompletely understood. In living organisms, glycation occurs naturally and has various implications, especially in the context of diseases like diabetes, but the full spectrum of its effects in biological systems has not been comprehensively studied.

In diabetic individuals, hyperglycemia-mediated glycation affects vascular function throughout the body⁴. Glycation-induced stiffness impairs blood vessels, and disturbs blood flow and nutrient delivery⁵. In the context of breast tumors, this impairment is even more critical. Diabetes leads to aberrant tumor vasculature characterized by irregularities in structure and function ⁶. Changes to the microvasculature can exacerbate cancer progression by facilitating tumor cell intravasation and metastasis^{7, 8}. Furthermore, the altered vasculature in diabetic breast tumors poses significant challenges to effective cancer treatment by hindering effective drug delivery⁹. Thus it is imperative to understand and address these diabetes-induced vascular changes for the development of more effective cancer therapies in patients.

Angiogenesis is modulated by an interplay of mechanical and chemical stimuli. Advanced Glycation End-products (AGEs), the product of glycation reaction, play a crucial role in this context by initiating cell signaling through their receptors (RAGEs) impacting the regulation of endothelial barrier function⁶. Our previous studies have established that increased matrix stiffness induces a tumor angiogenic phenotype and impairs endothelial barrier integrity^5,10. Specifically, endothelial monolayers adhered to stiff substrates exhibit increased cell contractility, leading to expanded cell-cell junctions and increased permeability. Additionally, the regulatory role of chemical cues in angiogenesis has been extensively studied, revealing a delicate balance between pro-angiogenic and anti-angiogenic factors¹¹. Pro-angiogenic factors initiate and promote new blood vessel formation. Conversely, anti-angiogenic factors inhibit endothelial cell proliferation and induce apoptosis. GM-CSF, a pro-angiogenic factor, promotes endothelial cell proliferation and migration, upregulates matrix metalloproteinases, and activates the JAK2/STAT-3 pathway¹². NF-κB, a downstream effector of both AGE-RAGE and Rho-kinase signaling, is instrumental in upregulating pro-inflammatory factors, including GM-CSF^1,13. Given the linkage between matrix stiffness, AGE-RAGE signaling, Rho-kinase mediated cell contractility, and NF-κB’s role in regulating tumor cell cytokine secretion, we investigated how tumor vasculature barrier function is regulated by glycation mediated NF-KB-mediated tumor cell cytokine secretion.

Here, we demonstrate diabetic hyperglycemia impairs tumor vasculature barrier function via glycation. Increased stiffness of extracellular matrix (ECM) and enhanced production of the AGEs due to glycation significantly increase cell contractility, activate NF-κB signaling, and promote GM-CSF secretion in tumor cells. This increased GM-CSF further impairs endothelial barrier function. Our data provides compelling evidence that hyperglycemia-induced glycation increases endothelial barrier function by increasing NF-κB regulated cytokine secretion in tumor cells. This insight into the glycation-mediated modulation of tumor microenvironment advances our understanding of the role of hyperglycemia in cancer progression and highlights potential targets for therapeutic intervention in oncology.

Results

Hyperglycemia promotes glycation-mediated AGEs production and matrix stiffening

Sugars are known to drive glycation reactions in vitro and in vivo^3,14. To investigate whether diabetic hyperglycemia triggers glycation reactions in tumor stroma, we employed a previously described STZ-based tumor-bearing diabetic MMTV-PyMT mouse model¹⁵. At 12 weeks of age, breast tumors from control and diabetic mice were harvested, lysed, and analyzed through western blotting. We observed a significant increase in AGEs production in tumors from STZ-treated mice compared to controls (Fig. 1a,b). This finding was corroborated by immunohistochemistry (IHC) assessments, where staining for AGEs in tumor sections aligned with the western blotting results, indicating larger AGE-positive areas in diabetic tumors (Fig. 1c,d). Intriguingly, diabetic mice treated with the glycation inhibitor aminoguanidine (AG) or the AGE-breaker ALT711 exhibited a significant reduction in AGE levels within breast tumor stroma, compared to untreated diabetic tumors. These data indicate that hyperglycemia enhances non-enzymatic glycation within the tumor stroma, thereby increasing AGEs production.

Figure 1. Hyperglycemia promotes glycation, which produces AGEs and increases matrix stiffness in vitro and in vivo.

a,b Protein bands generated by western blot and corresponding quantification showing AGEs in breast tumors extracted from MMTV-PyMT mice treated with (STZ) or without (Ctrl) STZ (N=6, n=6). c. Representative images of tumor sections with IHC staining for AGEs collected from non-diabetic (Ctrl), diabetic (STZ), and diabetic mice treated with glycation inhibitors (AG or ALT711). d. Corresponding quantification of normalized AGEs positive area (N=4, n=8). e. Equilibrium modulus of tumors extracted from control, diabetic, or diabetic mice treated with AG or ALT711 measured by AFM (N=3, n=3, 430-589 measurements per condition). f. Equilibrium modulus of collagen matrix glycated with various amounts of ribose measured by confined compression assay (N=4, 8 gels per condition). g. Equilibrium modulus of collagen matrix glycated with glucose measured by confined compression assay (N=4, 8 gels per condition). Data are presented as mean ± SEM. * p < 0.05, ** p< 0.01, *** p < 0.001, **** p < 0.0001.

Given that glycation reactions increase collagen matrix stiffness, we next investigated the effect of hyperglycemia-mediated glycation on the tumor stroma stiffness^3,16. Utilizing atomic force microscopy, we observed a significant increase in the ECM stiffness of diabetic tumors compared to controls (Fig. 1e). Notably, inhibiting glycation in diabetic mice reduced ECM stiffness of diabetic tumors to levels comparable to control tumors (Fig. 1e). While ribose has been extensively used in tissue engineering to stiffen collagen matrices in vitro, the effect of glucose is less clear. Thus, we glycated type I collagen matrices with varying concentrations of ribose or glucose and assessed matrix stiffness using a confined compression assay. Aligning with previous findings, our results showed that ribose increased collagen matrix stiffness in a concentration-dependent manner (Fig. 1f)³. Collagen matrices glycated with glucose show a similar increase in stiffness with rising glucose concentrations (Fig. 1g). These findings suggest that glucose-mediated glycation enhances the stiffness of collagen matrices and tumor stroma, in vitro and in vivo.

Hyperglycemia-mediated glycation reaction impairs tumor endothelial barrier function

Our prior data indicate that glycation of collagen matrix increases angiogenesis and impairs barrier integrity^5,10. Thus, we next investigated whether hyperglycemia-induced glycation influences angiogenesis and vascular function in breast tumors. Comparing vasculature density in control and diabetic tumors via CD31 staining revealed no significant differences, suggesting hyperglycemia does not alter the formation of new blood vessels within breast tumors (Fig. 2a,b). However, we detected a pronounced increase in vessel permeability in diabetic tumors when assessing vessel permeability via Evans Blue extravasation from 2MDa FIITC-dextran labeled vessels (Fig. 2c). As there were no significant differences in vessel density, our study concentrated on barrier function. To determine the effect of glycation on endothelial barrier function, diabetic mice were treated with the glycation inhibitors AG and ALT711. These treatments significantly improved endothelial barrier function in diabetic tumors, indicating hyperglycemia-mediated glycation damages tumor vasculature (Fig. 2c).

Figure 2. Glycation-induced matrix stiffening and AGEs production impair tumor endothelial barrier function.

a. Representative images showing vasculature in control and diabetic tumors. b. Quantification of vasculature density of tumors extracted from control and STZ-treated mice (N=4, n=8). c. Permeability of vasculature within control, diabetic tumors, or diabetic tumors treated AG or ALT711 (N=4, n=8). d. Quantification of permeability of vessel grown into collagen onplants with or without MDA-MB-231 in CAMs (N =3, 15 ~ 18 onplants per condition). e. Relative permeability of HUVEC monolayers treated with conditioned medium collected from MDA-MB-231 cells seeded on PA gels with 1 kPa or 10kPa stiffness and coated with glycated or unglycated collagen (N=3, 40-50 images per condition). Data are presented as mean ± SEM. NS, not significant, ** p < 0.01, *** p < 0.001, **** p < 0.0001.

To further investigate underlying mechanisms, we employed the CAM model and PA hydrogel system. In the CAM model, MDA-MB-231 cells were embedded in either glycated or unglycated collagen onplants. The permeability of vessels in these collagen onplants was compared by co-injecting 65–85 kDa TRITC-dextran and 2 MDa FITC-dextran, revealing higher permeability in vessels within glycated collagen, that was exacerbated when cancer cells were present (Fig. 2d). This indicates that glycation impairs vascular barrier function, an effect amplified by tumor cells.

Building on our previous work demonstrating the direct impact of ECM stiffness on endothelial barrier function, we investigated how tumor cells contribute to the further increase of permeability of endothelial cells in glycated microenvironments¹⁰. Considering the role of cell communication via signaling molecules and miRNAs, we cultured human umbilical vein endothelial cells (HUVECs) on PA gels, and treated them with conditioned medium from MDA-MB-231 cells cultured on varying stiffness substrates coated with glycated or unglycated collagen^17,18. We observed significantly increased permeability in HUVECs exposed to conditioned medium from cancer cells on stiffer, substrates coated with glycated collagen (Fig. 2e). These results collectively demonstrate that hyperglycemia-mediated glycation impairs vascular barrier function both in vitro and in vivo, with the presence of tumor cells exacerbating the detrimental effects of glycation on endothelial barrier function.

Glycation-induced ECM stiffening and AGEs formation shift tumor cell angiogenic secretome

To identify the mechanism by which tumor cell secretions in response to glycation result in impaired endothelial barrier function, we conducted bulk RNA sequencing on control, diabetic, and diabetic tumors treated with ALT711. We developed angiogenesis and cytokine scores to quantify the angiogenic phenotype and cytokine secretion status of tumors in each group. Our analysis revealed significantly higher angiogenesis and cytokine scores in diabetic tumors, indicating that hyperglycemia-mediated glycation enhances tumor angiogenic phenotype and the secretion of angiogenesis-related cytokines (Fig. 3 a,b). To further examine pathways influenced by glycation, we conducted Kyoto Encyclopedia of Genes and Genome (KEGG) pathway enrichment analysis. Diabetic tumors showed upregulation of cytokine–cytokine receptor interaction, TGF-β signaling, and Hippo signaling pathways, suggesting that hyperglycemia promotes inflammation, epithelial–mesenchymal transition, and activation of YAP/TAZ signaling. Additionally, the enrichment of the basal cell carcinoma pathway highlights increased oncogenic signaling in hyperglycemic tumors. Importantly, these pathway enrichments were largely diminished in diabetic tumors treated with ALT-711, suggesting that the observed alterations are primarily driven by glycation-induced AGE–RAGE engagement and matrix stiffening. As expected, AGE–RAGE signaling was selectively reduced in ALT-711–treated tumors (Fig. S1). To identify specific cytokines influenced by glycation-mediated AGEs and matrix stiffening, a cytokine array was performed, quantifying the secretion of angiogenesis-related cytokines in MDA-MB-231 cells cultured on compliant (1 kPa) or stiff (10 kPa) PA gels, treated with or without AGEs. This revealed that both AGEs and increased matrix stiffness affect the expression of pro-angiogenic factors, notably Granulocyte/Macrophage Colony-Stimulating Factor (GM-CSF), as evidenced by significant upregulation (Fig. 3c). This finding was further substantiated using ELISA (Fig. 3d).

Figure 3. Glycation-induced matrix stiffening and AGEs production shift tumor cell angiogenic secretome.

a. Angiogenesis score of control, diabetic, and diabetic tumors treated with ALT711 generated from bulk RNA sequencing (N = 3). b. Cytokine score of control, diabetic, and diabetic tumors treated with ALT711 generated from bulk RNA sequencing (N = 3). c. Angiogenic cytokine profiles of MDA-MB-231 cells seeded on compliant (1 kPa) or stiff (10 kPa) substrates and treated with or without AGEs (N=3). d. ELISA showing GM-CSF secretion of MDA-MB-231 seeded on 1 or 10 kPa PA gels and treated with or without AGEs (N =3). e. ELISA showing GM-CSF secretion of control, shCtrl, shRAGE cells, or cells treated with ZM1 seeded on 1 or 10 kPa PA gels (N=3). f. ELISA of GM-CSF secretion of cells on 1 kPa or 10 kPa PA gels and treated with or without ZM1 and/or Y27632 (N=3). Data are presented as mean ± SEM. * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001.

To dissect the contributions of matrix stiffness and AGEs to GM-CSF secretion in MDA-MB-231 cells, we employed two strategies to inhibit AGE-RAGE signaling: using the RAGE inhibitor FPS-ZM1 (ZM-1) and suppressing RAGE expression via shRNA. Both approaches markedly reduced GM-CSF secretion (Fig. 3e). Subsequently, we investigated the impact of matrix stiffness on GM-CSF secretion by inhibiting Rho/ROCK contractility pathway using Y27632¹⁹. ELISA measurements indicated a significant decrease in GM-CSF secretion upon Rho/ROCK pathway inhibition, an effect that was further pronounced when combined with the RAGE inhibitor ZM1 (Fig. 3f). These results suggest that both matrix stiffness and AGEs drive GM-CSF secretion, mediated through a combination of RAGE receptors and Rho/ROCK signaling pathways.

Glycation-induced matrix stiffening and AGEs production activate NF-κB signaling via contractility

Cell mechanosensing is dependent on cellular force generation, and AGEs have been shown to enhance contractility by interacting with mDia1 and activating Rho-mediated actin filament formation^20,21. The interplay between AGE-RAGE signaling, contractility, and mechanosensing, combined with our observations of increased angiogenic factor secretion in glycated matrices, suggests that glycation regulates GM-CSF secretion through contractility. To investigate this, we analyzed the contractility of MDA-MB-231 cancer cells exposed to AGEs, and treated with ZM1 or Y27632. Utilizing quantitative polarization microscopy, we detected increased cell contractility on stiffer substrates, which was significantly diminished upon inhibiting contractility or AGE-RAGE signaling (Fig. 4a)¹⁹. Furthermore, given the correlation between the phosphorylation of myosin light chain and contractility, we examined myosin light chain phosphorylation in MDA-MB-231 cells on compliant and stiff substrates, treated with AGEs or both AGEs and ZM1, via Western Blotting^20,22. The results consistently indicated that inhibiting AGE-RAGE signaling or cell contractility reduces myosin light chain phosphorylation (Fig. 4b,c).

Figure 4. Glycation-induced matrix stiffening and AGEs production activate NF-κB via contractility.

a. Average retardance of MDA-MB-231 cells on PA gels with different stiffness and treated with Y27632 (Y27) or ZM1 (N=3, 40-50 images per condition). b. and c. Western blot for phosphorylation of myosin light chain of MDA-MB-231 cells on PA gels with different stiffness and treated with Y27632 (Y27) or ZM1 (N=4). d. Confocal images of cells on PA gels with different stiffness, treated with Y27632 (Y27) or ZM1. Cells were stained for NF-κB and nucleus. e. Corresponding quantification of NF-κB signal within nucleus (N = 3; 40–50 imaging fields per condition). f. ELISA measurement of GM-CSF secreted by MDA-MB-231 cells seeded on PA gels and treated with an NF-κB inhibitor, JSH-23 (N=3). Data are presented as mean ± SEM. * p < 0.05, ** p<0.01, **** p < 0.0001.

Considering that NF-κB is a critical downstream effector in both AGE-RAGE and Rho-kinase signaling pathways, and its role in promoting the expression of multiple cytokines including GM-CSF, we further explored how AGE-RAGE signaling and cell contractility influence NF-κB activation and subsequent GM-CSF secretion^{1,13, 23}. MDA-MB-231 cells, seeded on PA gels of varying stiffnesses and exposed to AGEs, were treated with Y27632 or ZM1. NF-κB activation was quantified by measuring nuclear accumulation of NF-κB p65. We found increased NF-κB activation in cells on stiffer substrates, while inhibiting contractility or AGE-RAGE signaling reduced this activation (Fig. 4d,e). To confirm the generalizability of these findings, we validated this finding using the MCF-7 human breast cancer cell line. Briefly, MCF-7 cells were cultured on PA gels with stiffnesses of 1 kPa or 10 kPa and treated with or without AGEs. NF-κB activation was similarly assessed by NF-κB p65 nuclear localization. Consistent with our observations in MDA-MB-231 cells, NF-κB activation increased in MCF-7 cells cultured on stiffer substrates and following AGE treatment, indicating that glycation-induced matrix stiffening and AGE accumulation promote NF-κB activation across multiple breast cancer cell types (Fig. S2a,b).

To determine if glycation influences GM-CSF secretion in tumor cells via NF-κB, MDA-MB-231 cells were seeded on PA gels and treated with JSH-23²⁴. JSH-23 is an NF-κB inhibitor that specifically blocks nuclear translocation of NF-κB, thereby preventing downstream transcriptional activation²⁵. ELISA results revealed that GM-CSF secretion was higher in MDA-MB-231 cells on stiffer substrates, but was notably decreased upon NF-κB inhibition (Fig. 4f). We also repeated the experiment using MCF-7 cells and observed that GM-CSF secretion increased with substrate stiffness. Inhibition of NF-κB significantly reduced GM-CSF secretion in cells cultured on 10 kPa substrates. Although a decrease was also observed in cells on 1 kPa gels treated with JSH-23 compared to untreated controls on the same stiffness, the difference was not statistically significant (Fig. S2c). These findings collectively indicate that glycation-induced matrix stiffening and AGE production activate NF-κB through mechanisms involving contractility.

Glycation-mediated NF-κB activation regulates endothelial barrier function via GM-CSF secretion

Noting that glycation-induced matrix stiffening and AGE production activate NF-κB, thereby enhancing GM-CSF secretion, we aimed to elucidate the impact of this pathway on endothelial barrier function. We treated HUVEC monolayers with conditioned medium from MDA-MB-231 cells on PA gels of varying stiffnesses, and treated with or without JSH-23 to inhibit NF-κB. Interestingly, HUVECs fed with medium from JSH-23-treated tumor cells exhibited significantly reduced permeability (Fig. 5a). This underscores the role of NF-κB in modulating the effects of glycation on tumor cell angiogenic cytokine secretion, which in turn influences the endothelial barrier function. To further substantiate the role of NF-κB-regulated tumor cell GM-CSF secretion in affecting endothelial barrier function, HUVECs were treated with conditioned medium from MDA-MB-231 cells treated with shRNA for GM-CSF, cultured on compliant or stiff substrates. Consistent with prior observations, the permeability of HUVEC monolayers increased when treated with medium from MDA-MB-231 cells on stiffer substrates. However, the permeability significantly declined when HUVECs were treated with conditioned medium from MDA-MB-231 cells with diminished GM-CSF secretion (Fig. 5b). This finding suggests that matrix stiffening and AGE-induced tumor cell GM-CSF secretion detrimentally affect endothelial barrier function. Confirming our in vitro results, we conducted experiments using our diabetic breast tumor mouse model, assessing NF-κB activation and GM-CSF expression in tumor cells from diabetic and non-diabetic tumors. Diabetic tumors showed markedly higher NF-κB activation and GM-CSF secretion compared to their non-diabetic counterparts (Fig. 5c–f).

Figure 5. NF-κB-mediated GM-CSF secretion regulates endothelial barrier function.

a. Relative permeability of HUVEC monolayers treated with conditioned medium collected from MDA-MB-231 cells on compliant or stiff PA gels and treated with or without NF-κB inhibitor, JSH-23 (N=4, n=8). b. Relative permeability of HUVEC monolayers treated with conditioned medium collected from control or shGM-CSF cells seeded on PA gel substrates (N=4, n=8). c. Confocal images of tumors extracted from control or diabetic tumors. Tissue sections were stained for NF-κB and nucleus. d. Corresponding quantification of the percentage of cells with NF-κB⁺ nuclei within diabetic and non-diabetic tumors (N=3, n=4). e. and f. Representative images and corresponding quantification of GM-CSF expression within diabetic and non-diabetic tumors (N=3, n=4). Data are presented as mean ± SEM. * p < 0.05.

In summary, our research demonstrates that diabetic hyperglycemia increases the presence of AGEs in the matrix which disrupts tumor vasculature barrier integrity through glycation-mediated ECM stiffening and AGE production. Stiffened ECM and activated AGE-RAGE signaling lead to phosphorylation of myosin light chain, activation of NF-κB signaling, increased GM-CSF secretion in tumor cells, and impaired tumor vasculature barrier function.

Discussion

Diabetic hyperglycemia has been implicated in promoting tumor progression, with hyperglycemia-induced glycation as a major contributing factor¹⁵. However, the specific impact of hyperglycemia-mediated glycation on each hallmark of breast tumor progression, particularly tumor angiogenesis, remains under-explored. We present direct evidence that hyperglycemia-induced glycation compromises tumor vasculature barrier function. We delineate a pathway wherein increased tumor ECM stiffness and AGE-RAGE signaling consequentially activate NF-κB, elevate GM-CSF secretion in tumor cells, and impair the endothelial barrier function. Our findings demonstrate that hyperglycemia results in glycation, which elevates ECM stiffness and AGE concentration within the tumor stroma, subsequently increasing cell contractility and activating NF-κB. This glycation-mediated NF-κB activation alters the angiogenic cytokine profile of tumor cells and increases GM-CSF secretion. The resultant increases in GM-CSF secretion from tumor cells further deteriorates endothelial barrier function. Therefore, our research underscores the crucial role of glycation in regulating tumor vasculature barrier integrity and reveals that glycation-induced matrix stiffening, coupled with AGE production, regulates of tumor cell angiogenic cytokine secretion. Given the critical role of endothelial barrier function in tumor cell intravasation, a key step in tumor metastasis, our study also proposes a potential mechanism by which diabetes contributes to breast tumor metastasis through modulating tumor ECM²⁶. An important consideration not fully addressed in this study are the in vitro and in vivo impacts on barrier function compared to vessel density. While the impact of glycation on vascular barrier function was pronounced in vitro, changes in vascular density were less evident in vivo. Since multiple interacting factors modulate vascular responses and there are effectors that often promote both increased blood vessel growth and increased permeability, future research should aim to delineate the effectors of each in response to glycation.

In the biomaterials domain, extensive studies on glycation reactions have focused on collagen matrix stiffening without altering key mechanical properties such as pore size and fiber diameter^2,3. However, the biological roles of AGEs, products of glycation, remain elusive. Particularly the interaction of AGEs with cell mechanical sensing signaling remains incompletely understood. Our investigations reveal that glycation-driven AGE production initiates AGE-RAGE signaling, promotes the phosphorylation of myosin light chain, and augmentes cell contractility. Notably, matrix stiffening increased cell contractility and myosin light chain phosphorylation as well²⁷. This suggests a potential cooperative mechanism between glycation-induced matrix stiffening and AGE-RAGE in orchestrating cellular signaling cascades. Further, our data indicates a critical role of cell contractility in cell response to AGE-RAGE signaling. While our focus was primarily on the downstream impacts of glycation-induced AGEs production, it is pertinent to acknowledge the diversity within the AGE protein family²⁸. Future research delineating the specific AGE variants produced by glycation and their distinct biological functionalities is essential. Additionally, in this study, we utilized aminoguanidine and ALT-711 to inhibit glycation and assess its impact on tumor progression. While aminoguanidine and ALT-711 are two of the most commonly used glycation inhibitors with separate mechanisms of action, these inhibitors are known to have off-target effects, which may confound interpretation of our results^29–31. Therefore, the development of more selective and precise glycation inhibitors is necessary to better delineate the specific role of glycation in the regulation of tumor vasculature function.

Building upon our prior findings that matrix stiffness directly compromises endothelial barrier function, this study explores the contribution of tumor cells. We discovered that glycation shifts the tumor cell angiogenic cytokine profile, consequently altering endothelial cell behavior. Cytokines play an important role in cell-cell communication within the tumor microenvironment. Our data reveal that AGEs and ECM mechanical properties modulate intercellular communication across various cell types in the tumor microenvironment by altering cytokine secretion profiles. Significantly, tumor-associated macrophages, key components of the tumor immune microenvironment, also express RAGE, suggesting that the mechanisms we have uncovered may extend to other cell types³². Our findings corroborate the concept that the mechanical properties of the ECM, coupled with chemical cues in the tumor stroma, orchestrate the behavior of individual cell types and the communications among diverse cell populations, thereby facilitating a dynamic interplay among tumor cells, vasculature, immune cells, and fibroblasts³³. Notably, the role of GM-CSF may serve as a critical link between metabolic disorder-induced AGEs accumulation and alterations in the tumor immune landscape^34,35. The therapeutic potential of these interactions is underscored by recent advancements, such as the development of hydrogels co-loaded with CAR-T cells and GM-CSF to enhance T cell function, thus opening new avenues for treatment modalities³⁶.

NF-κB is a key regulator of cell-cell interactions and angiogenesis within microenvironments³⁷. While existing research predominantly focuses on how chemical cues induce NF-κB activation and subsequent angiogenesis, our findings introduce a novel perspective. We demonstrate that mechanical properties, notably matrix stiffness, can modulate NF-κB activity, thereby influencing cell-cell communication. Additionally, NF-κB’s role extends beyond angiogenesis, acting as a key regulator of inflammation^38,39. Thus, our results imply that matrix stiffness may also contribute to shaping the tumor immune microenvironment. This is further supported by our previous studies, which revealed that tumor cells on a stiffer matrix recruit more tumor-associated macrophages through elevated secretion of CSF-1⁴⁰. Collectively, these findings highlight the significance of mechanical properties in regulating the tumor immune microenvironment. While our study focused on glycation-induced NF-κB activation via contractility, it is important to note that the regulation of NF-κB by AGE–RAGE signaling is a complex process involving multiple signaling pathways, including NADPH oxidase, Ras/MEK/ERK1/2, SAPK/JNK, MAPK/p38, PI3K/AKT, small GTPases (Cdc42/Rac1), JAK/STAT, and GSK-3β pathways^41–44. Future studies will be needed to dissect how these signaling networks, downstream of AGE–RAGE engagement and ECM stiffening, collectively modulate NF-κB activity in the tumor microenvironment.

The relevance of NF-κB extends beyond breast tumors, as it is implicated in the progression of various diseases, including atherosclerosis, glioma, etc⁴⁵. Recent research reports that in atherosclerotic conditions, endothelial cells upregulate ICAM-1 and VCAM-1 expression through Jnk-dependent NF-κB activation, and alter their phenotype⁴⁰. Our prior studies have shown that atherosclerosis-induced increases in matrix stiffness impact endothelial barrier function⁴⁶. Therefore, our current study suggests the potential broader implications of matrix stiffness on NF-κB activity, suggesting the potential contribution of matrix stiffness to the progression of diseases beyond breast cancer.

During diabetes, AGEs are considered as a potential risk for β islet cell injury and insulin resistance⁴⁷. The activation of AGE-RAGE pathways increase reactive oxygen species production, enhancing mitochondrial superoxide production, thereby facilitating the development of diabetic nephropathy⁴⁸. In the context of breast tumors, RAGE activated by S100A7 proteins promotes tumor growth and metastasis via modulating ERK1/2, MAPK, and NF-κB signaling pathways in tumor cells and interaction between tumor cells and cancer-associated fibroblasts^38,49. Considering the crucial role of AGEs in both diabetes and cancer, substantial efforts have been directed toward developing a variety of AGE-inhibiting drugs^50,51. For instance, the RAGE antagonist FPS-ZM1 has demonstrated efficacy in protecting against glomerular filtration barrier damage in streptozotocin-induced diabetic rats⁵². Our findings reveal that inhibiting AGEs restores barrier function of tumor vasculature, suggesting that glycation inhibitors may enhance the efficacy of cancer therapies by normalizing vascular function. Cancer patients with pre-existing diabetes may benefit from integrating AGE-inhibiting treatments into traditional therapeutic strategies⁵³. There is the need for further investigation into the interplay between breast cancer and diabetes and the potential for other AGE-inhibiting drugs to have complementary effects. Moreover, the roles of angiogenesis and endothelial barrier function are critical in various other pathologies, including atherosclerosis, coronary artery disease, kidney fibrosis, and premature aging, suggesting that glycation inhibition might offer therapeutic benefits in these conditions as well^54,55.

Methods

Antibodies and reagents.

MDA-MB-231 breast adenocarcinoma cell (HTB-26) and MCF-7 human breast-cancer cell (HTB-22) were obtained from ATCC (Rockville, MD). Human Umbilical Vein Endothelial Cells (HUVECs) and endothelium growth medium (EGM) SingleQuot Kit were purchased from Lonza (Basel, Switzerland). DMEM media, penicillin-streptomycin, 0.25% trypsin, and trypsin neutralizer were purchased from Thermo Fisher Scientific (Waltham, MA, USA). Antibodies used for staining in western blot, immunofluorescence, and immunohistochemistry experiments were as follows: anti-AGEs antibody (ab23722, Abcam, Cambridge, UK), anti–glyceraldehyde-3-phosphate dehydrogenase (GAPDH) antibody (Poly6314; BioLegend, San Diego, CA, USA), anti-NF-κB p65 antibody (8242, Cell Signaling Danvers, MA, USA), anti-GM-CSF (ab9741, Abcam), anti-phospho-Myosin Light Chain 2 (Thr18/Ser19) Antibody (3674, Cell Signaling), anti- Myosin Light Chain 2 antibody (3672, Cell Signaling), anti–polyoma virus medium T antigen antibody (PyMT; ab15085, Abcam), anti-CD31 (1506-R, Santa Cruz, Dallas, TX, USA), anti-rat Alexa Fluor 647 goat antibody (A21247, Thermo Fisher Scientific), and anti-rabbit Alexa Fluor 568 donkey antibody (A10042, Thermo Fisher Scientific), Alexa Fluor 488 goat anti-rabbit antibody (A11008; Thermo Fisher Scientific), Advanced Glycation Endproduct-BSA (121800), NF-κB inhibitor JSH-23 (J4455), RAGE Antagonist FPS-ZM1 (553030), and ROCK inhibitor Y27632 (Y0503) were purchased from Millipore Sigma (Burlington, MA, USA). Acrylamide [40% (w/v) solution], bisacrylamide [2% (w/v) solution], ammonium persulfate, and tetramethylethylenediamine were purchased from Bio-Rad (Hercules, CA, USA). Alagebrium (ALT711) and alagebrium (AG) were purchased from MedChemExpress (Monmouth Junction, NJ, USA). All other reagents used here were purchased from Thermo Fisher Scientific.

Cell culture.

Highly metastatic MDA-MB-231 breast adenocarcinoma cells were maintained in DMEM supplemented with 10% FBS (Atlanta Biologicals, Flowery Branch, GA, USA) and 1% penicillin-streptomycin. MCF-7 human breast-cancer cells were maintained in Minimum Essential Medium (MEM;11095-080) supplemented with 10 % fetal bovine serum and 1 % penicillin–streptomycin. Medium was renewed every 2–3 days. For sub-culturing or downstream assays, MDA-MB-231 and MCF-7 were detached with 0.25 % trypsin–EDTA for ~2 min at 37 °C, neutralized with complete medium, and pelleted by centrifugation (1000 ×RPM, 5 min). HUVECs between p3 and p5 were cultured in EBM medium supplemented with the EGM SingleQuot Kit. Medium was replaced every 48 hours. Subculture was conducted using trypsin and trypsin neutralizer before confluence. To produce lentivirus particles, HEK293T cells were cultured in DMEM supplemented with 10% FBS and 1% penicillin-streptomycin. All cells were cultured at 37°C and 5 % CO2. All cell lines were tested and found negative for mycoplasma contamination using a universal mycoplasma detection kit (30-1012K^™, ATCC). To inhibit AGE-RAGE signaling, cells were treated with 1 μM PSF-ZM1⁵⁶. For ROCK inhibition studies, cells were treated with 10 μM Y27632 ¹⁹. To inhibit NF-κB signaling, cells were treated with 20 μM JSH-23²⁴.

shRNA-mediated knockdown of GM-CSF and RAGE.

For GM-CSF knockdown experiments, MDA-MB-231 cells were transduced with either PLKO shRNA scrambled control plasmid (#1864, Addgene, Watertown, MA, USA) or shRNA oligonucleotide for GM-CSF (targeting sequence: 5′-AACCTGAGTAGAGACACTG-3′)²³. To RAGE knockdown cell line, MDA-MB-231 cells were transduced with either RAGE shRNA (sc-36374-SH, Santa Cruz Biotechnology) or a Scramble shRNA control (sc-108060, Santa Cruz Biotechnology). Lentivirus particles were produced by transfecting each plasmid together with lentiviral expression vectors and second-generation packing constructs (psPAX2 and pMD2.G) in TransIT-LT1 (Mirus, Madison, WI, USA) into HEK293T cells. Lentiviral particles were harvested, concentrated with Lenti-X Concentrator (TaKaRa, Mountain View, CA, USA), and then transduced into MDA-MB-231 cells in the presence of polybrene (Santa Cruz Biotechnology). After transduction, cells were selected with puromycin. Knockdown efficiency was confirmed with western blotting.

Immunofluorescence.

MDA-MB-231 or MCF-7 on PA gels and tumor sections were fixed with 4% (v/v) paraformaldehyde, washed with PBS, and permeabilized with 1% (v/v) Triton X-100 in PBS. After fixation and permeabilization, all samples were blocked with 10% (v/v) FBS and 5% (v/v) donkey serum in PBS. Cells on PA gels were stained with primary antibodies (anti-NF-κB p65 antibody or anti-GM-CSF antibody) at 1:100 diluted in blocking solution overnight at 4°C, washed with PBS supplemented with 0.02% Tween, and then incubated with secondary antibody (donkey anti-rabbit Alexa Fluor 488) at 1:200 diluted in blocking solution for 1 h at room temperature in the dark. Tissue samples were stained following the same procedure with the primary antibodies at 1:50 dilution and secondary antibodies at 1:100 dilution. For anti-NF-κB activation assay, the nucleus was stained with DAPI. To outline tumor cells within tumor tissues, samples were also stained with PyMT antibody. Immunofluorescent images were taken with a Zeiss LSM800 microscope using an x40/1.1 N.A. water immersion objective and 488 excitation laser line. Protein colocalization analysis was performed as described previously¹⁰. Only PyMT+ cells were considered as tumor cells and included in NF-κB activation analysis.

Western blotting.

Snap-frozen tumors extracted from mice were ground using a pre-cooled mortar and pestle and dissolved in Laemmli buffer. The lysate was collected and spun at 14,000 xg at 4 °C for 10 min. The supernatant was collected and used for further analysis. For experiments quantifying protein expression, cells cultured on PA gels treated with PF673228 or PSF-ZM1 or DMSO before protein lysis. Total proteins were extracted with preheated 4x Laemmli buffer and stored at −80°C until use ¹⁰. Protein samples were subjected to gel electrophoresis [8.5 % (w/v) acrylamide gel for AGEs, 10 % (w/v) acrylamide gel for Myosin light chain] and western blotting. Membranes were blocked with either 5% milk or 5% BSA (MilliporeSigma) in Tris-buffered saline (TBS). Membranes were then incubated overnight with primary antibody at 4°C, and secondary antibody for 1 h at room temperature. Primary antibodies were prepared at 1:40,000 dilution in 5% milk in the case of GAPDH (MAB374), at 1:1000 dilution in 5% BSA in all other cases. Secondary antibodies were prepared at 1:2000 dilution. Membranes were imaged with West Pico, Dura, or Femto (Thermo Fisher Scientific) per their respective protocols using an ImageQuant LAS-4000 system in the other cases. Quantification was performed with ImageJ (National Institutes of Health). The relevant protein expression level was expressed as the ratio of the protein of interest to GAPDH. Phosphorlation level of myosin light chain was quantified by dividing the signal intensity of the phosphorylated myosin light chain by the total expression of the myosin light chain.

Atomic force microscopy.

Breast tumor tissues were collected from non-diabetic, diabetic, and diabetic mice treated with AG or ALT711, snap frozen, and sectioned at 20μm thickness. Stiffness of tumor ECM was measured using contact mode atomic force microscopy (MFP-3D, Asylum Research, Goleta, CA). Tissue sections were stained for nuclei with Hoechst 33342 fluorescent stain (Thermo Fisher). The tissue area without Hoechst signal was considered as ECM and was probed with a silicon nitride cantilever having a nominal spring constant of 0.06 N/m and 5 μm diameter borosilicate glass particle (Novascan, Boone, IA, USA). The spring constant of each probe was measured before use. Samples were incubated in PBS supplemented with protease inhibitor cocktail (Thermo Fisher Scientific) during the measurement to avoid degradation. Force-displacement curves were obtained by indenting 2 or 3 force maps (10-by-10) on the tissue sections with an approach and retract speeds of 1 μm/s until reaching the maximum set force of 3 nN. The Young’s modulus obtained from indention was extracted by fitting force-displacement curves to the Hertz model, assuming a Poisson’s ratio of 0.5.

Confined compression.

Mechanical properties of collagen matrices glycated with different amounts of glucose or ribose were measured using a TA Electroforce Model 3100 (TA Instruments) as described previously¹⁵. Collagen gels were subjected to 15% strain with 3% stepwise displacements over five steps. Each step consisted of a 1-mm indentation with a 15-minute relaxation time. Using a custom MATLAB script, the equilibrium modulus was calculated from the slope of the stress-strain curves generated using a poroviscoelastic model.

Immunohistochemistry.

Tumor tissue sections collected from control, diabetic, or diabetic MMTV-PyMT mice treated with ALT711 or AG were fixed with 4% (v/v) paraformaldehyde in TBS for 10 min and incubated with 10% FBS and 10% goat serum in TBS for 2 hours at room temperature. After blocking, tissue sections were then incubated with primary antibodies overnight at 4°C, followed up by exposure to 0.3% H2O2 in TBS for 15 min at room temperature to oxidize endogenous peroxidases in the samples. Samples were then incubated with a secondary antibody, horseradish peroxidase–conjugated anti-rabbit goat antibody (611-103-122, Rockland Immunochemicals, Limerick, PA, USA) at 1:400 diluted in TBS with 10% (v/v) FBS and 10% goat serum for 1 hour at room temperature. Samples were then incubated with 3,3’-Diaminobenzidine (DAB) (Cell Signaling Technology) for signal development and with Mayer’s hematoxylin for counterstaining. After dehydration and mounting, tissue sections were then digitalized using the Leica SCN400 Slide Scanner (Deer Park, IL, USA). Signal intensity and percentage of the signal-positive area were quantified with QuPath.

Collagen-coated polyacrylamide gel.

Polyacrylamide (PA) substrates were prepared as described previously¹⁰. Briefly, PA gels with 1 kPa or 10 kPa stiffness were synthesized by mixing different ratios of acrylamide [40% (w/v) solution] and bis-acrylamide [2% (w/v) solution] in Milli-Q water and TEMED (Bio-Rad) in 0.25 M HEPES buffer at pH 6. Polymerization was initiated by adding Ammonium Persulfate. Substrates were functionalized with N-6-[(acryloyl)amido]hexanoic acid, succi-nimidyl ester (N6), and coated with 0.1 mg/mL type I collagen (Corning, Corning, NY, USA). Unreacted N6 was neutralized using 1:1000 ethanolamine in 50 mM HEPES solution at pH 8. PA gels were washed in PBS and stored at 4°C in PBS with 2% penicillin/streptomycin before use.

Nonenzymatic glycation of collagen.

Type I collagen was isolated from rat tail tendons and reconstituted in 0.1% sterile acetic acid as described previously³. To glycate collagen, 10 mg/mL collagen solutions in acetic acid were mixed with 0.3 M glucose or ribose to form solutions containing various amounts of ribose or glucose and incubated for 5 days at 4 °C. Glycated collagen solutions were then polymerized by adding 1N NaOH in 10×PBS, HEPES, and sodium bicarbonate to form 1.5 mg/mL collagen gels with 1×PBS, 25 mM HEPES, and 44 mM sodium bicarbonate. Collagen was then incubated at 37°C for 1 h.

Quantitative polarization microscopy.

Cell contractility of cells seeded on soft or stiff PA gel substrates and treated with Y27632 or ZM-1 was measured with quantitative polarization microscopy (qPOL) as described previously¹⁹. Images were acquired using an inverted Axiovert microscope equipped with a Zeiss Axiocam 506 color camera and a 20× polarization lens (N.A.=0.5) with 10° intervals of the polarizer rotation over a range of 0 to 180°. The polarized image sequences were processed with a custom MATLAB code to obtain an optical retardance map. The retardance signal proportional to cell contractility was quantified by measuring the average retardance over the cell area after background subtraction.

Monolayer permeability assay.

200,000 HUVECs suspended in the EGM medium were seeded onto PA gels coated with 0.1 mg/mL type I collagen and cultured for five days to form confluent monolayers, with the medium replaced every 48 hours. Condition medium treatment and permeability assays were performed 2 days after cells formed a confluent monolayer. Conditioned media collected from MDA-MB-231 cells seeded on PA gels with different stiffnesses, or MDA-MB-231 cells treated with JSH-23 or MDA-MB-231 cells transduced with shRNA against GM-CSF were given to HUVEC monolayers, respectively. 40 kDa FITC-dextran solution (0.4 mg/mL) was prepared in medium. For each PA gel, 4 mL of this solution was added to a MatTek glass-bottomed dish (Ashland, MA, USA), all of which were stored in an incubator at 37°C at 5% CO₂. Seeded PA gels were transferred to filled dishes and visualized using a Zeiss LSM800 microscope equipped with a Plan-Apochromat ×20/0.8 N.A. objective (Carl Zeiss GmbH, Jena, Germany). Five minutes after submersion in FITC-dextran solution, 6 vertical line scans were captured on each gel spanning from 200 μm below the monolayer to 200 μm above the monolayer. Relative permeability of HUVEC monolayers was quantified by calculating the normalized intensity ratio (NIR) of each sample. Using ImageJ software, two rectangular boxes of the same size were drawn directly above and below the HUVEC monolayer. The average signal intensity within these two boxes was measured separately. The intensity ratio was calculated by dividing the average intensity within the gel by the average intensity above the gel. To produce NIR values, the intensity ratio values of seeded PA gels were divided by the intensity ratio values of unseeded control PA gels. NIR values of each sample were calculated by averaging the NIR of 8 vertical line scans that were captured on the same gel, as described previously¹⁰.

Chicken embryo culture.

Fertilized Leghorn chicken eggs were cleaned and stored in a wine cooler at 12.8°C on embryonic day (ED) 0 before transfer to a GQF 1500 Professional rocking egg incubator maintained at 37.8°C, 60% relative humidity. On ED3, eggs were cracked, and egg contents were transferred into sterile weighing boats capped with Petri dishes. A small scoop of autoclaved ground eggshell was distributed around the albumin. Chicks within culture platforms were incubated in Hova-Bator circulated air incubators (GQF, Savannah, GA, USA) maintained at 37.8°C, with 55% relative humidity.

Ex ovo neovessel permeability assay.

On ED10, collagen constructs were placed on the chick chorioallantoic membrane of chicken embryos. Briefly, unglycated and glycated collagen solutions loaded with or without MDA-MB-231 cells were polymerized for 30 min at 37°C, 5% CO₂ onto autoclaved nylon mesh squares as described previously¹⁰. On ED15, Proximal vessels were injected with 100 μL of mixture of 2.5 mg/mL 65–85 kDa tetramethylrhodamine (TRITC)-dextran (T1162, MilliporeSigma) and 2.5 mg/mL 2 MDa FITC-dextran (FD2000S, MilliporeSigma) in PBS. Twenty minutes after injection, neovessels within collagen constructs were visualized using a Zeiss LSM 800 upright confocal microscope equipped with W Plan-Apochromat 20x/1.0 N.A. water immersion objective. Neovessels were labeled by FITC-dextran. To quantify permeability of neovessels in collagen constructs, signal intensity of both the FITC-dextran and TRITC along a line eight pixels wide and perpendicular to the blood vessels were measured. The extravasated TRITC signal was then normalized to the TRITC present at the center of the vessel^5,10.

Enzyme-linked immunosorbent assay (ELISA).

MDA-MB-231 cells or MCF-7 cells were cultured on PA gels with 1kPa or 10kPa stiffness coated with 0.1 mg/mL Type 1 rat tail collagen. After culturing overnight, cell culture medium was replaced with DMEM supplemented with 0.5% FBS and 1% penicillin-streptomycin. After 24-hour culturing, cell culture medium was collected and subjected to ELISA. For inhibition studies, all cells were fed with AGEs. For each condition, cells were treated with DMSO, or ZM1, or Y27632, or JSH-23. For RAGE inhibition study, conditioned medium were collected from MDA-MB-231 cells treated with ZM-1 or transduced with shRNA targeting GM-CSF. Conditioned medium and GM-CSF standard were then analyzed using Human GM-CSF DuoSet ELISA kit (DY215, R&D system, , Minneapolis, MN, USA). Samples were dissolved in the coating buffer (DY006, R&D system) and loaded into 96-well microtiter plates (896036, R&D) respectively. Plates were incubated overnight at 4°C, washed three times with washing buffer (WA126, R&D), and blocked with reagent dilute (841380, R&D) for 2 hours at room temperature. After blocking, plates were washed and incubated with rabbit anti-GM-CSF antibody for 2 hours at room temperature. Wells were then washed and incubated for 2 hours with HRP-conjugated igG. After incubating with secondary antibody, plates were then washed, and signal was developed by adding a 1 : 1 mixture of reagent A and B (895000, 895001, R&D) and incubating at room temperature for 20 min. After signal development, the stop reagent (895926, R&D) was added to each well. The optical density (OD) of samples was measured using a microplate reader set to 450 nm. A standard curve was generated using OD collected from the GM-CSF standard. AGEs concentration of each sample was calculated using this standard curve.

Cytokine profile.

MDA-MB-231 cells were cultured on PA gels with 1 kPa or 10 kPa stiffness coated with 0.1 mg/mL Type 1 rat tail collagen treated with or without AGEs. After culturing overnight, cell culture medium was replaced with DMEM supplemented with 0.5% FBS and 1% penicillin-streptomycin, After 24-hour culturing, cell culture medium was collected and utilized as directed by the Proteome Profiler Human Angiogenesis Array Kit (ARY007, R&D Systems).

MMTV-PyMT transgenic mice studies.

FVB/N-Tg(MMTV-PyVT)634Mul/J (MMTV- PyMT) female mice were utilized in this study. All mice were maintained following a protocol approved by the Vanderbilt University Institutional Animal Care and Use Committee (Protocol # M1700029-01). Female MMTV-PyMT mice of the FVB strain background (Jackson Laboratory, Bar Harbor, ME, USA) were fed with D12492 high-fat diet (NC0004611, research diet Inc, New Brunswck, NJ, USA) starting at 4 weeks of age. To induce diabetes, mice were given 5 consecutive daily doses of streptozotocin (STZ, MilliporeSigma) through intraperitoneal injection at 70 mg/kg body weight as described previously¹⁵. The control group of mice were injected with citrate buffer (0.1 mmol/L, pH 4.5; MilliporeSigma) as vehicle control. For STZ-injected mice receiving aminoguanidine, mice were treated with 3 mg/kg body weight aminoguanidine (MilliporeSigma) in drinking water until the study endpoint. For STZ-injected mice receiving alagebrium (ALT711) treatment, mice were treated with 1 mg/kg body weight alagebrium (MedChemExpress, Monmouth Junction, NJ, USA) injected daily intraperitoneally until study endpoint. Seven weeks after STZ injection, permeability of tumor vasculature was measured as described previously^5,10. A saline solution of Evans blue (50 mg/kg body weight) and 2-MDa FITC-dextran (40 mg/kg body weight) was injected retroorbitally, and the dyes were allowed to circulate for 1 h. The mice were humanely sacrificed by CO₂ asphyxiation and necropsied. The mammary tumors were collected and snap frozen, and embedded in optimum cutting temperature (O.C.T.) compound and were processed in a cryostat to obtain 20-μm-thick sections, and mounted in Vectashield mounting medium (Vector Laboratories, Newark, CA, USA). Fluorescent images were acquired using a 10×/0.3 N.A. objective on a Zeiss LSM 800 laser-scanning microscope. Permeability quantification was performed by measuring the signal intensity of both the FITC-dextran and Evans blue dyes along a line perpendicular to the blood vessels. The center of each vessel was found by fitting the FITC-dextran dye signal with a Gaussian curve. The extravasated Evans blue signal then was normalized to the Evans blue present at the center of the vessel. To quantify vascular density, mammary tumors were collected and snap frozen and then sectioned at 5 μm thickness for immunohistochemical staining and at 20 μm for atomic force microscopy (AFM).

RNA isolation from MMTV-PyMT tumors.

Upon sacrifice, tumors were surgically removed from the mice, and small pieces of the tumors were cut and snap-frozen in liquid nitrogen. RNA was isolated from snap-frozen tumor sections using a combination of TRIZOL (#15596026, Thermo Fisher Scientific) and RNeasy micro kits (#74004, Qiagen, Hilden, Germany). In brief, tumor sections were combined with 5 mm stainless steel beads (69989, Qiagen) in 1 mL of ice-cold TRIZOL into 2 mL microcentrifuge tubes. The samples were then homogenized by placing the tubes in a TissueLyser II device (#85300, Qiagen) set to shake at 30 Hz for 2 minutes. After ensuring samples were completely homogenized, the lysate was transferred to a new 1.5 mL tube without the stainless steel beads. Samples were allowed to incubate at room temperature in TRIZOL for 5 minutes. 0.2 mL of chloroform was added to each sample, shaken vigorously, and allowed to incubate at room temperature for 2-3 minutes. The samples were centrifuged at 4°C for 30 minutes at 12,000 xg. The aqueous top layer was carefully removed and added to a new tube containing 0.5 mL 70% ethanol and mixed carefully by pipetting up and down several times. The RNA and ethanol mix was applied to the RNeasy spin columns via centrifugation at 10,000 xg for 30 seconds. RNA and ethanol mix was applied to the columns in 0.5 mL aliquots until all samples had been applied to the column. The samples were washed with 0.7 mL RW1 buffer and centrifuged at 10,000 xg for 30 seconds. The flow-through was discarded, and samples were washed with 0.5 mL RPE buffer and centrifuged at 10,000 xg for 30 seconds. The flow-through was discarded, and samples were washed with 0.5 mL RPE buffer and centrifuged at 10,000 xg for 2 minutes. The flow through was discarded, and the columns were placed into fresh collection tubes and centrifuged at 10,000 xg for an additional 1 minute. Then RNA was eluted in 35 μL of RNase-free water and centrifuged at max speed (14,000 xg) for 1 minute in fresh 1.5 mL microcentrifuge tubes. RNA concentration and purity were measured via nanodrop. Samples were stored at −80°C until further use.

RNA sequencing and analysis.

RNA samples were submitted to the Vanderbilt Technologies for Advanced Genomics (VANTAGE) core facility for bulk RNA sequencing. Samples underwent quality control via RNA ScreenTape to ensure all samples had a RINe greater than 7. All of our samples had RINe numbers greater than 8.1, and most were greater than 9.0. The VANTAGE core performed mRNA enrichment and cDNA library preparation via the stranded mRNA Library Prep kit (New England Biolabs, Ipswich, MA, USA). Paired-End 150 bp sequencing was performed on the Illumina NovaSeq 6000 targeting an average of 50M reads per sample. Demultiplexed FASTQ files obtained from the VANTAGE core were aligned to the Mus musculus transcriptome using Kallisto. To build the kallisto transcriptome index, we utilized the Mus_musculus.GRCm38.cdna.all.fa.gz file found at ‘https://ftp.ensembl.org/pub/release-96/fasta/mus_musculus/cdna/’ and the Mus_musculus.GRCm38.9.gtf.gz file found at ‘https://ftp.ensembl.org/pub/release-96/gtf/mus_musculus/’. Differential expression analysis was carried out in R using DEseq2. Outputs from Kallisto runs for every file and analysis scripts in the form of jupyter notebook files (.ipynb) were deposited in the Open Science Framework (OSF) and is accessible at DOI: 10.17605/OSF.IO/GQK9P. Angiogenic and cytokine scores were computed by taking the average z-score for manually curated lists of genes relevant to angiogenesis and cytokines. Genes included in the calculation of angiogenic and cytokine scores are listed in Supplemental Information (Table S1). To identify pathways altered by hyperglycemia-mediated glycation, KEGG pathway enrichment analysis was performed. DEGs from bulk RNA (|log₂FC| > 1, FDR-adjusted p < 0.05) from each pair-wise contrast (Control vs STZ, and STZ vs STZ + ALT) were analyzed using the enrichKEGG function in the clusterProfiler package in R. Only pathways with at least 3 differentially expressed genes (DEGs) were retained for downstream analysis. The entire set of genes tested by DESeq2 was used as the background. For visualization, the five most significant pathways overall (lowest adjusted p values, irrespective of direction) were plotted for each comparison. For the Ctrl vs STZ comparison, pathways were selected to illustrate both biologically relevant changes in pathways associated with cancer-related signaling and the most significantly enriched pathway in the opposite direction, to ensure representation of changes across both upregulated and downregulated gene sets. For STZ vs STZ + ALT comparison, the five most significant pathways overall (lowest adjusted p values, irrespective of direction) were plotted for each comparison.

Statistical analysis.

Statistical analysis was performed using GraphPad Prism 10 (GraphPad Software, La Jolla, CA, USA). Data here are presented as means ± SEM. T test was used when comparing two groups. When more than two groups were analyzed, parametric one- or two-tailed one-way analysis of variance (ANOVA) followed by a Mann-Whitney test or a post hoc Tukey’s test were used where appropriate. P < 0.05 was considered statistically significant.

Resource availability

Lead contact

Further information and requests for resources and materials should be directed to the lead contact, Dr. Cynthia Reinhart-King (cr73@rice.edu).

Materials availability

This study did not generate new unique materials.

Data and code availability

All data reported in this paper will be shared by the lead contact upon request. RNA sequencing data is deposited at Gene Expression Omnibus (GEO) with accession number: GSE285474 and are publicly available as of the date of publication. The code used for RNA sequencing analysis has been deposited in the Open Science Framework (OSF) at DOI: 10.17605/OSF.IO/GQK9P and in Zenodo at DOI: 10.5281/zenodo.15777581.

Supplementary Material

Document S1. Table S1, and Figures S1–S2.

Bigger Picture statement.

While glycation is a well-accepted approach to stiffening collagen for tissue engineering applications, its physiological effects in vivo remain less understood. Here, we show that diabetic hyperglycemia impairs tumor vasculature integrity via glycation. Glycation-mediated ECM stiffening and AGEs formation increase cell contractility, NF-κB activity, and GM-CSF secretion in tumor cells, subsequently impairing tumor vessels. The mechanism we report has major implications for understanding how diabetes affects tumor progression, emphasizing the previously unrecognized role of glycation in the link between diabetes and breast cancer. Additionally, clinical evidence suggests that glycation plays a key role in aging and the progression of various diseases, such as neurodegenerative diseases, atherosclerosis, and heart failure. Meanwhile, NF-κB affects the progression of various diseases like glioma. Our study shows that glycation may play a broader role in disease progression beyond cancer.

Highlights.

• Diabetic hyperglycemia enhances glycation reaction within breast tumor ECM.

• Glycation alters tumor cell angiogenic secretome.

• Glycation activates GM-CSF secretion via contractility-mediated NF-κB activation.

• Elevated GM-CSF secretion damages tumor vasculature barrier function.