TGFβ enhances platelet-breast-cancer-cell interaction and promotes platelet aggregation.

Introduction
Breast cancer (BC) is the most commonly diagnosed malignancy in women, with over 2 million new cases reported annually [1]. The tumor microenvironment in BC can be examined at multiple levels: local (intra‐tumoral), regional (within the breast), and distant (metastatic), each comprising a variety of cell types [2]. Among these, platelets (PLT) have gained increasing attention due to their significant role in tumor progression [3]. PLT have been found to have a key role in both BC dissemination and disease progression [4]. Elevated PLT levels have been correlated with worse progression‐free survival and overall survival in BC [5, 6].
PLT are highly dynamic secretory cells that modify their microenvironment by releasing bioactive molecules (such as growth factors, chemokines, clotting factors, RNAs, and extracellular vesicles) in response to external stimuli [7]. Beyond hemostasis, PLT are involved in several processes, such as inflammation, immune response, angiogenesis, and tumor progression [3, 8]. Accumulating evidence suggests that cancer cells interact with PLT, directly or indirectly, leveraging their protective and pro‐metastatic properties to facilitate tumor progression [3, 4, 8, 9, 10].
PLT contribute to both primary tumor growth and metastasis by infiltrating the tumor microenvironment and supporting tumor cells in epithelial–mesenchymal transition (EMT), chemoresistance, and intravasation [4, 11]. Additionally, circulating tumor cells (CTC) have been shown to recruit PLT in the bloodstream, serving multiple functions, including shear stress protection, immune evasion, as well as enhancing survival and facilitating extravasation [3, 4, 8, 9, 10].
Notably, cancer cells can “educate” PLT to further support tumor progression (tumor‐educated platelets, TEP) by transferring RNAs and proteins. Moreover, tumor cells can induce PLT activation and aggregation (tumor‐induced platelet aggregation, TCIPA) through the release of soluble agonists or via direct cell‐PLT interactions [3, 9, 12, 13], ultimately enhancing their protective role in circulation [4, 9, 13, 14].
TCIPA, along with patient comorbidities and cancer treatments, contributes to a hypercoagulable state known as cancer‐associated thrombosis (CAT), a leading cause of cancer‐related mortality [13, 15]. Altered levels of tissue factor (TF), thrombin, adhesion receptors, and other soluble factors have been observed in cancer patients, despite their direct association with CAT remaining elusive and largely unexplored [16]. Among the growth factors, transforming growth factor beta (TGFβ) has been hypothesized to be implicated in both PLT‐induced aggregation and cancer progression [13, 17].
TGFβ, which is actively secreted upon PLT activation, promotes tumor progression by enhancing migration, invasion, proliferation, angiogenesis, EMT, and cancer stem cell maintenance [18]. PLT‐derived TGFβ has been demonstrated to facilitate metastasis via TGFβ/Smad and NF‐κB signaling pathways [19]. Additionally, other studies have shown that MCF7 BC cells interact with PLT via integrins (e.g., α2β1) [20, 21], triggering TGFβ autocrine signaling that promotes EMT, invasion [20], and PLT aggregation [14]. However, the interplay between TGFβ signaling, PLT–cancer cell interactions, and cancer cell‐induced PLT aggregation remains unclear. Here, we investigated whether TGFβ pre‐stimulated BC cells (MCF7), before PLT addition in culture, showed increased propensity to interact with the latter, focusing on both PLT‐MCF7 adhesion and MCF7‐induced PLT aggregation. MCF7 have been selected as a model for this study since they retain an epithelial phenotype [22] and they are known to undergo EMT upon TGFβ stimulation [23, 24].

Results

TGFβ stimulated EMT in MCF7 cells
Since TGFβ is known to induce a more aggressive phenotype [23, 24] in MCF7, we investigated the effect of TGFβ (10 ng·mL−1, [23, 25, 26, 27, 28]) on MCF7 proliferation, migration, and EMT. Both TGFβ[29, 30] and cell–cell adhesion [29, 30, 31] regulate function and expression of cell adhesion proteins (e.g., integrins), which play in turn also a key role in PLT‐cells [6] interaction. To mitigate the potential confounding effects of cell culture density, MCF7 treatment with TGFβ has been conducted under two distinct culture densities: 20 000 cells per cm2 (low plating density) and 60 000 cells per cm2 (high plating density) (Fig. 1A). These conditions were selected to evaluate potential density‐dependent influences on PLT‐MCF7 interactions following 48‐h stimulation with TGFβ, since a differential effect of TGFβ stimulation was previously proved in cellular models, including MCF7 [27, 32].
We found that cell proliferation, as determined by the WST‐1 assay, was unaffected by TGFβ treatment in both cell culture densities (Fig. 1B). We also determined MCF7 migration capacity in the different conditions with a scratch assay. As described in the Materials and Methods, we seeded MCF7 at the different densities and treated them with TGFβ for 48 h. Afterwards, we re‐seeded the cells from all the conditions tested with the same density in order to obtain a fully confluent layer at 12 h, and we performed the scratch (this time point has to be considered the 0 h), which has been followed for a further 48–72 h. Through this experimental design, we found that TGFβ promoted a higher migration capability in cells at low plating density 48 h after the scratch was executed, while no effect was observed on high plating density cells (Fig. 1C). To assess the EMT, we first investigated cellular circularity, which reflects cellular elongation that occurs during the mesenchymal transformation [23]. TGFβ treatment was observed to significantly stimulate MCF7 stretching both at 20 000 cells per cm2 and 60 000 cells per cm2, while the different plating densities did not affect cellular morphology in both the NT and TGFβ conditions (Fig. 1D).
To confirm the observed phenotypic change attributable to EMT, we analyzed two markers of EMT previously found to be dysregulated in TGFβ‐treated MCF7 [23]: E‐cadherin and SNAI1. We first measured the exposure of E‐cadherin on the cell surface, finding a significant decrease after TGFβ treatment compared to NT at both plating densities. Moreover, at high plating density, E‐cadherin was downregulated in both conditions (Fig. 1E). We also found that SNAI1, VIM, and TWIST1 gene expression were upregulated in low plating density cells treated with TGFβ compared to the other conditions (Fig. 1F). Collectively, these results confirmed that TGFβ treatment promoted EMT of MCF7, while some of the EMT hallmarks tested (scratch closure and SNAI1 expression) were abolished by increasing cellular culture density.

PLT‐MCF7 interaction in suspension was promoted by TGFβ
Since PLT and CTC are known to interact in the blood flow [6, 33, 34], we investigated PLT‐MCF7 interplay in suspension after adherent MCF7 were pre‐stimulated for 48 h with TGFβ (Fig. 2A). PLT labeled with CFSE and incubated with MCF7 were analyzed by flow cytometry, as summarized in Fig. 2A. MCF7 positive for CFSE fluorescence (CFSE+) were selected in order to identify the cells that interacted with PLT, as described in Materials and Methods.
Through this approach, we initially investigated the effect of cell culture density on the interaction between PLT and MCF7 cells treated with TGFβ (Fig. 2B). Cells treated with TGFβ at a culture density of 20 000 cells per cm2 showed a significant increase (P < 0.05) in the capacity to bind PLT when compared to the control group (untreated cells, represented by the blue bars) and when compared to both 60 000 cells per cm2 conditions. At high plating density, no significant difference was observed between the TGFβ‐treated and untreated samples, suggesting an enhanced effect of TGFβ with reduced cell density. Representative contour plots of TGFβ−/+ PLT‐MCF7 interaction are shown in Fig. 2C.
To investigate whether TGFβ contained in PLT releasate similarly promoted PLT‐MCF7 interaction, we pre‐treated MCF7 at 20 000 cells per cm2 for 48 h with PLT releasate; this plating density was selected since we observed an effect caused by TGFβ stimulation. The latter was prepared as described in the Materials and Methods and found to be 1.6 μm ± 0.9, as expected from the literature [35, 36]. From this analysis, we found that the treatment of MCF7 cells with platelet releasate resulted in a significant increase of PLT adhesion as compared to the untreated condition (NT), similarly to the effect observed with TGFβ treatment when compared with the same NT (Fig. 2D).
The addition of an inhibitor of the TGFβ receptor (TGFβi) was able to restore the NT condition, confirming this finding. Our results showed that inhibition of this pathway significantly counteracted the ability of MCF7 cells to bind PLT, represented by the yellow bar (P‐value < 0.05 for releasate + TGFβi vs. respectively releasate and TGFβ conditions). These findings suggest that supplementation with TGFβ enhances the ability of MCF7 cells to interact with PLT and that this effect can be mimicked by platelet‐derived releasate containing TGFβ. This finding suggests a functional role for platelet‐derived TGFβ in modulating tumor cell‐platelet interactions, which may contribute to tumor progression mechanisms within the circulation.

Platelets activation and aggregation were promoted by TGFβ‐treated MCF7
PLT and MCF7 were pre‐incubated in order to stimulate their interaction without triggering PLT aggregation, with 1.25 mm CaCl2 and 30% T‐PAS+.
To determine the effect of MCF7 (grown at 20 000 cells per cm2) pre‐treated with TGFβ on PLT activation, we analyzed the surface expression of the PLT activation marker CD62P (P‐selectin), as previously described [36] (Fig. 3A). This plating density was selected since we observed an effect of TGFβ stimulation. Both NT and TGFβ‐treated MCF7 significantly stimulated PLT activation (P‐values < 0.001), with a significant increase in CD62+ PLT in the TGFβ condition compared to the NT (Fig. 3B, P‐value < 0.05). This result suggests that PLT‐MCF7 interaction fosters PLT activation and that this effect is further promoted by TGFβ pre‐treatment of MCF7. Histograms of CD62 positivity are reported in Fig. 3C.
The effect of TGFβ‐treated or not treated MCF7 (grown at 20 000 cells per cm2) on the induction of PLT aggregation was investigated following PLT activation with thrombin through a spectrophotometric assay [37, 38]. As shown in Fig. 4A, we examined changes in light transmittance over 30 min (min) to monitor PLT aggregation [37, 39, 40]; maximum aggregation was measured at 30 min. Unstimulated PLT were also included for comparison as negative controls (− Thrombin, Fig. 3D). The aggregation slopes in the presence of PLT activated with thrombin showed that mean aggregation values of TGFβ are higher than NT, and the Δ between the two slopes increases over time with a higher final maximum aggregation at 30 min. Conversely, the aggregation slopes of both NT and TGFβ conditions incubated with non‐activated PLT were almost flat, with a slight increase only after 15 min for the TGFβ condition (Fig. 3D). Comparing the maximum aggregation values at 30 min, we found a significant increase in platelet aggregation in the TGFβ‐treated MCF7 cells compared to NT (P‐value< 0.01), while both conditions are significantly higher in the presence of activated PLT (Fig. 3E
P‐values< 0.001).

Integrin‐αν/CD51 and Galectin‐3 were upregulated in MCF7 upon stimulation with TGFβ
To investigate the mechanisms of PLT‐MCF7 interaction as dysregulated by TGFβ, we analyzed the gene expression of cell surface proteins known to be involved in PLT‐tumor cells interaction [6] (i.e., ADAM9, ITGAV, ITGB1, LGALS9, LGALS3, PDPN). The analysis was performed on MCF7 seeded at low and high plating density. The expression of ITGAV and LGALS3 was significantly higher in the TGFβ condition than in the NT (P < 0.05) at 20 000 cells per cm2, while we did not find significant differences at 60 000 cells per cm2 (Fig. 4A). Interestingly, ITGAV expression was upregulated in NT at high plating density while LGALS3 at 60 000 cells per cm2 was downregulated when compared to 20 000 cells per cm2 for both NT and TGFβ (Fig. 4B).
To further investigate the effect of TGFβ, we analyzed CD51/Integrin‐αν (ITGAV) and Galectin‐3 (LGALS3) exposure on the membrane surface of MCF7 with flow cytometry. Accordingly, we observed higher levels of both proteins in the TGFβ condition compared to the NT at low plating density, both in terms of mean fluorescence intensity (MFI) (Fig. 4C) and % of positive cells (Fig. 4D), whereas no increase was detected at 60 000 cells per cm2 (Fig. 4C,D). Representative histograms and contour plots are represented in Fig. 4E,F. Contrary to what was observed for RNA expression, there was no difference between the NT conditions at 20 000 cells per cm2 and 60 000 cells per cm2 when evaluating protein exposure, suggesting that TGFβ treatment may exert a different effect on gene expression depending on cellular culture density, but not on protein exposure.

Inhibition of integrin‐αν/CD51 reversed the increase in PLT‐MCF7 promoted by TGFβ interaction whereas inhibition of Galectin‐3 did not
To confirm the involvement of Galectin‐3 and Integrin‐αν/CD51 in PLT‐MCF7 interaction, we selected known inhibitors that specifically block their functions. The inhibitors tested included GB1107 [41] for Galectin‐3, Cilengitide [42], and GLPG0187 [43] for Integrin‐αν/CD51. The latter was found not to be specific for Integrins‐αν, but to also target α5β1 integrin [43]. Each compound was used at 1 μm, based on previous in vitro studies [41, 44, 45] (Fig. 5A). Inhibitors were added to MCF7, PLT, or their combination for the 45 min incubation of PLT‐MCF7 interaction assays. All compounds were first tested for cell viability to rule out any cytotoxic effects on MCF7.
Cilengitide, GB1107, and GLPG0187 were then assessed for their cellular toxicity and capability to inhibit TGFβ‐induced PLT‐MCF7 interaction using the experimental setting as described in Fig. 5B. None of the inhibitors resulted in being non‐toxic at our working concentration (1 μm) (Fig. 5C). Cilengitide restored PLT‐MCF7 interaction to the level of the NT condition (TGFβ vs Cilengitide P‐value < 0.05), while neither GB1107 nor GLPG0187 counteracted the enhanced interaction promoted by TGFβ (Fig. 5D).

Inhibitors of integrin‐αν/CD51 and Galectin‐3 counteracted the effect of TGFβ pre‐treatment on MCF7‐induced PLT aggregation
To determine the effect of Galectin‐3 and Integrin‐αν/CD51 inhibition on PLT‐MCF7 interaction, we had to exclude a possible involvement of these inhibitions on PLT aggregation capacity and, in case they were not, whether we can consider them as specific inhibitors of PLT–cancer cell interaction (Fig. 6A). From previously published studies we know that Integrin‐αν was excluded from being involved in PLT aggregation [46], while Galectin‐3 was previously proposed to have an active role in PLT aggregation, although its mechanism of action has not been fully elucidated [47, 48]. More specifically, Cilengitide was found not to inhibit PLT aggregation at our working concentration [49] while GLPG0187 and GB110 have never been tested before for PLT aggregation. We found that neither Cilengitide nor GB1107 affected PLT aggregation (Fig. 6B). Conversely, GLPG0187 almost abolished PLT aggregation capability (Fig. 6B); since neither αv Integrins nor α5β1 integrin [50] should be involved in PLT aggregation, this may be attributed to a non‐specific or unknown effect of GLPG0187.
All the inhibitors were then tested for their capability of interfering with the PLT‐MC7‐induced aggregation (Fig. 6C) and we found that they were able to reduce PLT aggregation in the TGFβ condition, restoring it to levels comparable to the NT (Fig. 6D,E). While the effect of GLPG0187 can be attributed to the inhibition of PLT aggregation itself (Fig. 6B,D and E), both GB1107 and Cilengitide's effect may be ascribed to a more specific effect on MCF7‐induced PLT aggregation (Fig. 6D,E). Interestingly, despite all the inhibitors reducing PLT aggregation (Fig. 6D,E), only Cilengitide was able to affect MCF7‐PLT interaction (Fig. 5D).

Analysis of BC proteomic datasets revealed a significant correlation between TGFβ1 and the platelet marker ITGA2B (CD41)
Although the presence of intra‐tumor PLT still remains largely unexplored [51, 52], PLT have been found in BC biopsy specimens [11] and their presence correlated with chemotherapy response but not with nuclear grade and BC stage. In this study, BC samples were not subdivided into the main BC subtypes (e.g., Luminal, Basal). Here we examined the correlation (Spearman correlation) between the protein levels of TGFβ1 and ITGA2B (CD41), the latter identified as a specific platelet marker [53], in a previously published proteomic dataset of BC, subdividing for BC subtypes and tumor advancement (by stage or lymph nodes involvement). The aim of this analysis was to identify possible enrichments of PLT presence in the tumor tissues that correlate with TGFβ.
In particular, we used protein expression data from a study by Johansson et al. [54] and Krug et al. [55]. As highlighted in the graphs (Fig. 7A,B), a significant correlation was observed between the expression of TGFβ1 and ITGA2B in the luminal A breast cancer tumor subtype of both datasets (respectively r = 0.883, P‐value = 0.003 and r = 0.302, P‐value = 0.033) and in triple‐negative breast cancer (TNBC) for the dataset from Krug et al. (r = 0.542, P‐value = 0.005). These findings support the usage of MCF7 as a valid in vitro model and suggest a possible heterogeneity in PLT–BC interaction. Similarly, TGFβ1 and ITGA2B were found to correlate in BC with at least 1 lymph node involvement and not in BC without lymph node involvement (Fig. 7C, from the dataset of Johansson et al.), while in both stage II and III BC a significant positive correlation was found (Fig. 7D, from the dataset of Krug et al.).
To confirm that TGFβ may enhance PLT‐MCF7 in the tumor site, we tested PLT interaction with adherent MCF7. The interaction between the cells and CFSE‐labeled PLT was assessed by counting the % of CFSE+/DAPI cells with immunofluorescence microscopy (Fig. 7E).
Representative immunofluorescence images are illustrated in Fig. 7E, fields = 3 for n = 3 replicates counted. This analysis showed a significantly higher fluorescence intensity of PLT incubated with TGFβ‐treated cells compared to the NT (Bars in Fig. 6E, P‐value < 0.05). Altogether, these findings suggest that TGFβ may promote PLT‐BC cell interaction both in the tumor site and in the blood flow, eventually leading to increased PLT aggregation and potentially fostering augmented protection of the tumor cells (Fig. 7F).

Discussion
Breast cancer cells–PLT interaction has been extensively studied. PLT have been suggested to interact with circulating BC cells, providing protection from shear stress and immune surveillance, and facilitating metastatic seeding [5, 9, 10]. In particular, PLT have been identified at the primary tumor site, where they may contribute to cancer cell extravasation [11]. Cancer cells (including BC [14]) have been shown to promote PLT aggregation, also known as TCIPA [3, 9, 12, 13], which in turn increases the shielding capacity of PLT for circulating cancer cells and augments the risk of cancer‐associated thrombosis (CAT) [9].
TGFβ, a primary PLT effector, contributes with PLT to promote EMT and tumor aggressiveness [19, 20, 21].
The main objective of this study was to determine whether MCF7, pre‐stimulated with TGFβ and incubated with PLT, exhibited an increased propensity for interaction with PLT. Our finding confirmed that TGFβ‐treated MCF7 undergo EMT, a key feature of cancer aggressiveness. To mimic the interaction of PLT with circulating BC cells, MCF7 cells, pre‐stimulated with TGFβ or PLT releasate, were left to interact with PLT in suspension. Either TGFβ or PLT releasate stimulation of MCF7 promoted an increased propensity to bind PLT when compared to NT cells. In parallel, we observed that TGFβ‐treated MCF7 were able to elicit PLT activation and aggregation, as demonstrated by CD62 exposure and aggregation kinetics.
We hypothesized that this enhanced interaction was due to an increased expression and/or exposure of specific adhesion proteins on the cell surface, which in turn promoted PLT aggregation. To test this hypothesis, we analyzed the gene expression of cancer cell adhesion proteins involved in cancer cells‐PLT interaction [6]. Our findings revealed that Galectin‐3 (LGALS3) and CD51/Integrin‐αv (ITGAV) were upregulated upon TGFβ treatment at a cell density of 20 000 cells per cm2, leading to greater surface exposure of their protein products, thus confirming that TGFβ‐induced EMT acts through the regulation of cell–cell interacting proteins [56, 57, 58]. However, this effect was lost at a higher cell density (60 000 cells per cm2), suggesting that this mechanism is controlled by cell–cell junction formation [29, 30, 31] and by cellular culture density. Previous studies reported that an intact cellular monolayer was able to suppress TGFβ‐induced EMT, while this was reactivated at low plating density [27, 32]. This data support the hypothesis that cancer cells with an aggressive, transformed phenotype (enabling them to evade normal regulatory mechanism) may be more responsive to TGFβ stimulation and, consequently, exploit PLT interaction to foster cancer dissemination.
Next, we tested specific inhibitors of Galectin‐3 and CD51/Integrin‐αv to investigate their involvement in increased PLT aggregation as a consequence of the enhanced PLT‐MCF7 interaction and to exclude a fundamental involvement of soluble effectors. GLPG0187 (Integrin‐αv and α5β1 inhibitor) interfered directly with PLT aggregation, independent of MCF7 stimulation. Conversely, both GB1107 (Galectin‐3 inhibitor) and Cilengitide (Integrin αv inhibitor) did not reduce PLT aggregation itself, while they proved their efficacy in abolishing the increase in PLT aggregation promoted by MCF7 pre‐treated with TGFβ. Nevertheless, only the specific inhibition of Integrin αv mediated by Cilengitide was able to reduce PLT‐MCF7 interaction. These findings indicate that the interaction between PLT and cells, as well as cell‐induced PLT aggregation, is partially independent mechanisms. Both Galectin‐3 and Integrin‐αv cell adhesion proteins play a crucial and specific role in triggering cell‐mediated PLT aggregation upon TGFβ stimulation, while they appear to have a minor role in PLT‐MCF7 interaction.
Intriguingly, the inhibition of Integrin‐αν with Cilengitide reverted the increased PLT–MCF7 interaction, whereas GLPG0187 [43] did not. This discrepancy may be attributed to differences in their in vitro inhibitory efficacy, as prior studies have shown that Cilengitide exerts a stronger blockade of Integrin‐αv compared to GLPG0187 [59]. Additionally, Cilengitide may also interfere with Integrin αIIbβ3 on PLT [49] and, although this effect appeared to be weak in terms of PLT activation and aggregation (as confirmed here), it may exert a more specific influence on the interaction between MCF7 and PLT. Albeit the involvement of soluble factors produced by TGFβ‐treated MCF7 can be excluded, the potential contribution of MCF7‐derived extracellular vesicles (EVs) should be further investigated, since these EVs may exploit adhesion proteins to facilitate the interaction with PLT. Moreover, a role for EVs in CAT has already been demonstrated [15]. The importance of Galectin‐3 and Integrin αv in PLT–tumor interaction has been previously investigated, supporting the hypothesis of their involvement in the proposed mechanisms of PLT–MCF7 interaction. In particular, Galectin‐3 [60] and Integrin αvβ3/αvβ5 [61] were found to play a key role in EMT, cell invasion, and interaction with the microenvironment. Moreover, Galectins inhibition was found to be promising at both pre‐clinical and clinical levels, significantly reducing cancer metastasis and improving the therapeutic outcome [62]. Recently, GB1107 was also proved to decrease PLT‐stimulated gastric cancer invasion, both in vitro and in vivo [63], despite its direct effect on PLT–cell interaction and on PLT behavior not being investigated. Importantly, Galectin‐3 was proposed to have an important role in venous thrombosis [47] and has also been shown to mediate cancer cell–PLT interactions that promote cancer progression [64]; nevertheless, these associations were not investigated in CAT [47]. Intriguingly, inhibiting the Galectin‐3 ligand GPVI (specifically expressed on cancer cells) [6] reduced metastases spreading in mouse models [65]. Similarly, while the Integrin‐αv inhibitor Cilengitide has shown promising results in various clinical studies, inconsistent outcomes have limited its therapeutic usage [66]. In BC, Cilengitide was proposed to diminish bone metastasis [67, 68] and recent evidence suggests that its therapeutic efficacy may be affected by the different integrin profiles across BC subtypes [69].
The present study found that TGFβ induces adhesive proteins, thereby promoting PLT–cancer cells interaction and, consequently, PLT aggregation. In this context, PLT–cancer cells interaction may not only support cancer dissemination but also be a key mechanism underlying CAT. However, the precise mechanisms driving this process are not fully elucidated and, while TGFβ has been proposed as a potential mediator, its role has yet to be validated [13, 17]. We acknowledge that the platelet‐to‐tumor cell ratio used in our in vitro experiments does not reflect physiological conditions, where the number of circulating platelets vastly exceeds that of tumor cells. However, this experimental ratio (1 : 600) was selected based on similar studies on TCIPA in vitro [14, 20, 21, 70, 71, 72] and to allow quantifiable modulation of platelet–tumor cell adhesion. Therefore, this setup was chosen as a compromise between biological relevance and experimental feasibility. Moreover, the handling method and storage of PLT, including the type of anticoagulant agent and additive solution employed, have the potential to impact their functionality. It has been demonstrated that PLT stored within the first 2 days from collection and preserved in T‐PAS+ maintain their properties and respond to agonists [73, 74, 75, 76, 77]. Nevertheless, PLT possible alterations caused by preservation and handling may be a limitation for in vitro studies, and the results obtained necessitate further investigation in vivo.
While further in vivo studies are needed to validate this proposed association, our findings may pave the way for the development of more tailored treatments for CAT [78] and for counteracting metastatic dissemination [65]. Such strategies could be based on TGFβ, Galectin‐3, and Integrin‐αv inhibitors, given that current antiplatelet approaches increase bleeding risk and fail to reduce metastases [79].
Furthermore, our analysis of published proteomic datasets [54, 55] demonstrated that a positive correlation between TGFβ and a PLT marker (ITGA2B/CD41) occurs within the primary luminal A cancer tissue, the same subtype of our experimental cell model (MCF7). This finding is consistent with the previously reported presence of PLT in BC tissue [11] and is confirmed by our in vitro PLT‐MCF7 assessment in adhesion, where TGFβ‐treated MCF7 cells bound significantly more PLT than untreated NT cells. Similarly, PLT also showed a significant positive correlation with TNBC, though only in one of the proteomic datasets analyzed. The heterogeneity of this association across BC subtypes is consistent with recent findings that seem to suggest a specific cancer‐dependent PLT involvement [65, 69]. The TGFβ‐CD41 correlation was significant in both stage II and III cancers, but only when at least one lymph node was involved, suggesting that PLT may have a more prominent role in aggressive or advanced disease. This evidence is consistent with our finding, showing increased PLT‐MCF7 interaction upon TGFβ stimulation. In contrast to this observation, Ishikawa [11] did not find a significant correlation between PLT staining and BC stage in the analyzed surgical specimens, even though TGFβ involvement was not taken into account. Although the detection of CD41 (ITGA2B) in the proteomic dataset cannot unequivocally demonstrate platelet infiltration in breast cancer tissue, its expression, together with previous evidence of platelet presence in tumor samples, supports the hypothesis of platelet involvement in the tumor microenvironment, warranting further validation through spatial or histological approaches.
Taken together, our findings suggest that TGFβ may initially stimulate PLT‐BC cell interaction at the primary tumor site, facilitating extravasation and supporting cancer cells during the subsequent phases of cancer dissemination (as depicted in Fig. 7F). Further in vivo experiments are necessary to confirm this hypothesis and to better understand the mechanisms underlying this process. Future studies should also focus on cancer cells with more aggressive phenotypes and different cancer types to determine whether similar mechanisms are involved. Additionally, investigating soluble factors (such as mRNA, proteins, and extracellular vesicles) that MCF7 cells may transfer more efficiently following TGFβ stimulation could provide deeper insights into their contribution to tumor progression and microenvironment modulation. These efforts will contribute to a more comprehensive understanding of the interplay between cancer cells and their surroundings, potentially unveiling novel therapeutic targets.

Materials and methods

Cell culture and treatment conditions
MCF‐7 cell line (RRID:CVCL_0031) was purchased from LGC Standards s.r.l. Italy (part of LGC Limited, Teddington, Middlesex, UK). Cells were maintained at 37 °C, 5% CO2 (Midi CO2 Incubator; ThermoScientific, Waltham, MA, USA), grown in the standard condition with complete low glucose Dulbecco's modified Eagle's medium (DMEM Low Gluc; EuroClone, Milan, Italy) supplemented by 10% fetal bovine serum (FBS; EuroClone), 1% L‐Glutamine (EuroClone), and 1% Penicillin–Streptomycin (Pen‐Strep; Sigma Aldrich, St. Louis, MI, USA). The growth medium was renewed 2–3 times per week, and cell confluency was kept approximately under 80% in T25 flasks (CytoOne; Starlab, Milan, Germany). To ensure myco‐free results, a mycoplasma test (MycoAlert Mycoplasma Detection Kit; Lonza, Basel, Switzerland) was performed every 2 weeks according to the manufacturer's instructions during the whole experimental session.
The experimental protocol for testing the effect of TGFβ pre‐treatment of MCF7 phenotype and PLT‐MCF7 interaction was developed by seeding cells at 20 000 cells per cm2 or 60 000 cells per cm2 with complete DMEM low glucose medium. After 24 h of culture, the medium was replaced by adding complete DMEM low glucose (NT) or DMEM low glucose with TGFβ (T7039; Sigma Aldrich) at 10 ng·mL−1 for 48 h before each experimental assay. For conditions with Integrin‐αν and Galectin‐3 inhibitors, the following specific inhibitors were added to PLT and cells at 1 μm immediately before performing the specific assay: Integrin‐αν inhibitor Cilengitide resuspended in water (Tocris Bioscience, Bristol, UK), GLPG0187 (Tocris Bioscience) resuspended in DMSO, Galectin‐3 inhibitor GB1107 (DBA Italia, Italy) resuspended in DMSO. 1 : 1000 DMSO (Sigma‐Aldrich) was added in the non‐treated (NT) and TGFβ conditions experiments when necessary. PLT and MCF7 were incubated for 45 min for each experimental assay. Similarly, Integrin‐αν and Galectin‐3 were added to MCF7, PLT, or to the combination of both for 45 min. The experimental conditions are further described in the following paragraph and summarized in schematic representations within each figure in the results section.

Cell viability by WST‐1 assay
The WST‐1 colorimetric assay was performed as previously published [36]. MCF7 in the different conditions were cultured in a 96‐Well plate (Primo® multiwell plate; EuroClone) with complete DMEM low glucose medium. After 48 h of treatment, the medium was removed and cells were incubated with 100 μL of WST‐1 cell proliferation reagent (Merck, Darmstadt, Germany) for 2 h at 37 °C in a CO2 incubator prior reading at 450 nm (GloMax Discover; Promega, Madison, WI, USA).

Cell migration assay
MCF7 at 20 000 cells per cm2 or 60 000 cells per cm2 was seeded in a 6‐well plate (Primo® multiwell plate; EuroClone) with complete DMEM low glucose medium.
After 24 h of culture, the medium was replaced by adding DMEM low glucose (NT), or DMEM low glucose with TGFβ (10 ng·mL−1). After 48 h of culture in the NT and TGFβ (at 20 000 or 60 000 cells per cm2) conditions, cells were detached and re‐seeded into 24‐well plates at a density to allow 90% confluency at the standard growing condition and incubated overnight. 12 h later, a linear wound was made in the MCF7 monolayer using a sterile 10 μL pipette tip (ClearLine; Dominique Dutscher, Bernolsheim, France), the wells were washed twice with PBS to remove cellular debris and DMEM low glucose supplemented by 0.1% FBS was added. Digital images of the wound area were taken at 4× magnification immediately after wounding (0 h) and at subsequent time points (24 and 72 h after the scratch). Data were analyzed using ImageJ software [80] to measure wound closure or the number of migrated cells.

Analysis of cellular circularity
Morphological analysis of MCF7 cells was performed to evaluate the effect of TGFβ on cell circularity [81]. After 48 h of treatment, digital images were captured at 20× magnification and analyzed using ImageJ software to assess the effect of TGFβ treatment on cell circularity and polarization. The area and perimeter of 15 cells per condition were measured.
Circularity was calculated using the following formula:

Determination of MCF7 surface proteins using flow cytometry
To determine the level of E‐cadherin, CD51 (Integrin‐aν) and Galectin‐3 expressed on MCF7 surface upon TGFβ stimulation, MCF7 were grown in the different conditions as described above. After treatment, cells were detached, centrifuged at 450 
g
, and resuspended in PBS to obtain a concentration of 1 × 106 cells·mL−1.
For each experimental condition, 0.2 × 106 cells were labeled with 1 : 50 CD51 (CD51 Antibody, anti‐human, REAfinity, Miltenyi Biotec, Bergisch Gladbach, Germany) and Galectin‐3 (Galectin‐3 Antibody, anti‐human/mouse, REAfinity, Miltenyi Biotec) for 20 min at room temperature in the dark. For E‐cadherin, cells were incubated in PBS with a 1 : 100 anti‐E‐cadherin antibody (Ab 1416; Abcam, Cambridge, UK), and subsequently with a 1 : 400 secondary antibody (Goat anti‐mouse IgG Highly Cross‐Adsorbance Secondary Antibody, Alexa Fluor Plus 488; Invitrogen, Waltham, MA, USA) for 15 min at room temperature in the dark. Cells were then centrifuged at 450 
g
and resuspended in 400 μL PBS. Samples were analyzed using a FACSLyric flow cytometer (BD, Bioscience, East Rutherford, NJ, USA).

Gene expression analysis
MCF7 cells were cultured in a 12‐well plate at 20 000 or 60 000 cells per cm2. NT and TGFβ MCF7 grown for 48 h in the respective condition were analyzed for gene expression of EMT markers previously found to be dysregulated in TGFβ‐treated MCF7 [23, 24] and cell adhesion proteins previously found to be involved in PLT–cancer cell interaction [6].
RNA was extracted using MONARCH® Total RNA Miniprep Kit (New England Biolabs, Ipswich, MA, USA). RT was carried out on total RNA (0.500 μg) using RevertAid First Strand cDNA Synthesis kit (Thermo Fisher Scientific) and 2720 Thermal Cycler (Applied Biosystems, Waltham, MA, USA). Quantitative PCR (qPCR) was performed using GoTaq® qPCR Master Mix (Promega). The primer pairs used are shown in Table 1. qPCR was conducted using the CFX Duet Real‐Time PCR System (Bio‐Rad, Hercules, CA, USA). GAPDH was used as a housekeeping reference. ΔC
t and 2−ΔCt were calculated to quantify relative RNA expression.

PLT collection
The study was approved on 17 December 2020 (protocol number 2020/0149618) by the local Institutional Board Review (Reggio Emilia Ethics Committee). Twenty‐five Platelet donors were recruited between January 2022 and December 2023 by the Transfusion Medicine Unit of the Azienda USL‐IRCCS di Reggio Emilia upon signing informed consent. The study methodologies conformed to the standards set by the Declaration of Helsinki. PLT were obtained through the apheresis procedure as previously described [36] using an automated blood collection system (Mobile Collection System MCS+, Haemonetics Corp., Boston, MA, USA) in Acid Citrate Dextrose Solution A (ACD‐A, Haemonetics Corp.) and platelet additive solution (T‐PAS+; Terumo, Tokyo, Japan). Fresh PLT are stored at 22–24 °C with constant agitation for no more than 5 days and used herein 2 days from collection, and before each experiment, PLT count was performed on a 10 mL aliquot of fresh PLT using a blood cell counter (XS‐1000i analyzer; Sysmex, Kobe, Hyōgo Prefecture, Japan).

PLT Releasate
To assess the contribution of constitutive TGFβ released by PLT during the process of activation and degranulation [35], PLT releasate was prepared 1 day prior to cell treatment. PLT (1 × 109·mL−1) were centrifuged at 2000 
g
for 10 min. Pellets were resuspended with complete DMEM low glucose medium to the initial PLT concentration. Next, PLT were incubated for 10 min at 37 °C in presence of 0.62 U·mL−1 Thrombin (Merck KGaA) to induce activation and degranulation. PLT pellets were separated from the releasate (i.e., supernatant containing the molecules released after PLT degranulation) by centrifugation at 2800 
g
for 7 min. The supernatant was transferred into a new tube and 2 U·mL−1 Heparin was added. The releasate was filtered using a 0.22 μm sterile Syringe Filter (Corning®, Corning, NY, USA). The TGFβ in the releasate was quantified with ELISA assay (R&D Systems, Minneapolis, MN, USA) according to manufacturer guidelines.

PLT‐MCF7 interaction in suspension
For PLT labelling, 1 × 109 PLT·mL−1 were diluted 1 : 3 in Dulbecco's phosphate buffered saline (PBS, EuroClone) with the addition of 2 U·mL−1 Heparin (Pfizer, New York, USA) and 8 mm Ethylene glycol‐bis (2‐aminoethylether)‐N,N,N′,N′‐tetraacetic acid (EGTA; Merck) to avoid clotting; next, PLT were labeled with 5 mm 5(6)‐Carboxyfluorescein diacetate N‐succinimidyl ester (CFSE; Sigma Aldrich) for 10 min in agitation, protected from the light. Labeled PLT were centrifuged at 1350 
g
for 5 min, the supernatant was discarded, and the pellet was washed twice with T‐PAS+. PLT were resuspended at the initial concentration prior to incubation with MCF7 in binding medium, composed of DMEM Low Glucose, 0.1% FBS and 25 mm Calcium Chloride (Merck).
The interaction between MCF7 cells and PLT at the different conditions was performed in suspension using a modified version of a previously published method [21].
MCF7 were prepared as follows. MCF7 were initially seeded at different plating densities (20 000 or 60 000 cells per cm2). After 24 h of culture, the medium was replaced with the described experimental conditions (+/− TGFβ). After 48 h treatment, MCF7 were detached, centrifuged at 450 
g
, and resuspended in complete DMEM low glucose medium to obtain a concentration of 1 × 106 cells·mL−1. For the releasate experiments, cells at 20 000 cells per cm2 were grown with DMEM used to obtain the releasate (see the previous paragraph). TGFβ inhibitor (TGFβi; Sigma Aldrich SB431142) was resuspended in DMSO and used at a final concentration of 1 μm. An equal concentration of DMSO (1 : 1000) was added to each condition of the same experiment.
For each experimental condition, 0.2 × 106 cells were incubated with 120 × 106 CFSE‐labeled PLT (MCF7 : PLT = 1 : 600), and the final volume was adjusted to 500 μL of binding medium and incubated at 37 °C for 45 min in continuous agitation. For the conditions with integrin inhibitors, the specific integrin inhibitors were added to the PLT and cells. Successively, samples were further diluted with 400 μL of PBS before acquisition at the flow cytometer. As a gating strategy, PLT and MCF7 cells were firstly discriminated based on their physical parameters (SSC/FSC). Subsequently, MCF7 displaying CFSE positivity (CFSE+) were considered for interaction analysis.

PLT activation by MCF7 interaction
MCF7 treated for 48 h in the different conditions was incubated in suspension with PLT. Specifically, 0.2 × 106 cells were incubated with 120 × 106 PLT (MCF7 : PLT = 1 : 600) in binding medium as described before.
Next, MCF7‐PLT samples were incubated at room temperature with the activation marker anti‐P‐selectin fluorescent antibody (CD62P‐APC Miltenyi Biotec) in the dark for 15 min and then diluted 1 : 10 with TPAS before acquisition with the FACSLyric flow cytometer.

Microplate aggregation assay
Platelet aggregation was assessed in a 96‐well plate as previously described [37, 38]. For each experimental condition, 0.5 × 106 cells·mL−1 were incubated at 37 °C for 45 min in agitation with 300 × 106 PLT·mL−1 to obtain a ratio of MCF7 : PLT = 1 : 600.
For conditions with Integrin‐αv and Galectin‐3 inhibitors, the inhibitors were added together with platelets and cells as described before and in the schematic representations.
Cells and PLT were incubated in a solution containing 30% of T‐PAS+ and a final concentration of CaCl2 of 1.25 mm in order to stimulate MCF7–PLT interaction without stimulating aggregation, which was activated by the addition of Thrombin.
Before the experiment, PLT were centrifuged at 1350 
g
for 10 min and resuspended in 1 mL of T‐PAS+ solution. Cells were centrifuged at 450 
g
for 5 min and resuspended in 70 μL of PBS solution containing 1.80 mm CaCl2 (Farmalabor, Canosa di Puglia, Italy) and 1.05 mm MgCl2(Thermo Fisher Scientific). Subsequently, 100 μL of the cell‐platelet mixture was aliquoted in a 96‐well plate. 0.62 U·mL−1 Thrombin or an equal volume of PBS (in the non‐activated control) were added to each well. After PLT activation, the plate was maintained at 37 °C in cycles of 50 s rest and 10 s shaking. Absorbance was read at 405 nm every minute for 30 min in a plate reader (GloMax Discover; Promega, USA). As a direct measure of PLT aggregation, transmittance was calculated from the average of the light intensity measurements (A = absorbance) (%Transmittance = 100*10−A).

Cell viability by propidium iodide staining and flow cytometry
The toxicity of inhibitors on suspension of MCF7 incubated with PLT was assessed using the propidium iodide (PI) staining protocol according to the manufacturer's (Sigma‐Aldrich). For each experimental condition, cells and PLT were mixed as described for the aggregation assay.
The resuspended cells were incubated with PLT and 1 μm of the following inhibitors for 1 h in the Glomax: integrin Galectin‐3 inhibitor (GB1107), integrin‐αv inhibitor (Cilengitide), or αv and α5β1 Integrins inhibitor (GLPG0187). After the incubation, cell viability was analyzed using PI staining and assessed with a FACSLyric flow cytometer (BD Biosciences).

Correlation of TGFβ1 and ITGA2B expression in breast cancer subtypes proteomic dataset
We exploited the protein expression data from two literature studies (Johansson et al. [54] and Krug et al. [55]) to investigate the possible correlation between the expression of TGFβ1 and ITGA2B (CD41, a PLT‐specific surface marker) in different subtypes of BC. Using R Studio software (version 4.3.3), the data were processed and a Spearman correlation test was performed (GraphPad Prism 7.05 version). Scatter plots were generated for each BC subtype to visually represent the relationship between these two proteins.

PLT‐MCF7 interaction in adhesion
MCF7 cells (20 000 cells per cm2) were seeded on coverslips within the wells of a 24‐well plate (Primo® multiwell plate, EuroClone). After 24 h of culture, the medium was replaced by adding 1000 μL of DMEM low glucose (control group) and DMEM low glucose with TGFβ (10 ng·mL−1) for 48 h. After 48 h of treatments, media were removed and each well was incubated by adding 100 μL of PLT labeled with 5 mm CFSE, as previously described, with 400 μL complete DMEM low glucose and 2 U·mL−1 Heparin. The plate was agitated for 45 min at 37 °C.
After 45 min, the medium was gently removed from the wells and coverslips were washed in PBS three times and fixed with 1% paraformaldehyde for 10 min (Thermo Scientific).
Slides were incubated overnight at +4 °C with primary antibody, anti‐E‐cadherin (36/E‐Cadherin ab287970 Abcam) at the dilution of 1 : 100 in PBS. After 24 h, samples were washed four times with PBS and incubated at RT for 1 h in the dark with 1 : 1000 secondary antibody, Goat anti‐Mouse IgG (H + L) Alexa Fluor™ 568 (Thermo Fisher Scientific), stained and mounted using Fluoroshield with DAPI (Sigma Aldrich). Images were collected with a Leica DM6B Fluorescent Microscope (Leica, Wetzlar, Germany).

Statistical analysis
Statistical analysis was performed using Microsoft Excel (2019 version) and GraphPad (GraphPad Prism 7.05 version). The Student's t‐test was used to compare two samples, while one‐way ANOVA with multiple comparisons was used to compare more than two sets of samples. Comparisons were considered significant when (*P < 0.05, **P < 0.01, ***P < 0.001). Schematic representations were created with Biorender.

Conflicts of interest
The authors declare no conflict of interest.

Author contributions
MG, LM, RA, GG, CM, and DS performed experiments and collected data. MG, LM, CM, EM, and DS analyzed data. MG, CM, LM, EM, and DS wrote the original manuscript. LM, CM, and DS supervised experiments. RB acquired funding and supervised experiments.