Mammary Microvessels are Sensitive to Menstrual Cycle Sex Hormones

Abstract The mammary gland is a highly vascularized organ influenced by sex hormones including estrogen (E2) and progesterone (P4). Beyond whole‐organism studies in rodents or cell monocultures, hormonal effects on the breast microvasculature remain largely understudied. Recent methods to generate 3D microvessels on‐chip have enabled direct observation of complex vascular processes; however, these models often use non‐tissue‐specific cell types, such as human umbilical vein endothelial cells (HUVECs) and fibroblasts from various sources. Here, novel mammary‐specific microvessels are generated by coculturing primary breast endothelial cells and fibroblasts under optimized culture conditions. These microvessels are mechanosensitive (to interstitial flow) and require endothelial–stromal interactions to develop fully perfusable vessels. These mammary‐specific microvessels are also responsive to exogenous stimulation by sex hormones. When treated with combined E2 and P4, corresponding to the four phases of the menstrual cycle (period, follicular, ovular, and luteal), vascular remodeling and barrier function are altered in a phase‐dependent manner. The presence of high E2 (ovulation) promotes vascular growth and remodeling, corresponding to high depletion of proangiogenic factors, whereas high P4 concentrations (luteal) promote vascular regression. The effects of combined E2 and P4 hormones are not only dose‐dependent but also tissue‐specific, as are shown by similarly treating non‐tissue‐specific HUVEC microvessels.


Introduction
The microvasculature is essential for constructing physiologic organ-on-chip systems, as the presence of blood vessels is Vascular heterogeneity exists between different organs and tissues, where metabolic demands are supported by different types of microvasculature. [7]Although phenotypic vascular differences are known to exist across tissues, the intrinsic genetic and epigenetic signatures promoting specialized organ-specific vasculature are still poorly understood, [8] but largely influenced by the microenvironment. [9]2c,10] Recent studies show the influence of tissue-specific fibroblasts on the transcriptome of endothelial cells, giving rise to different subtypes of cells with unique gene expression profiles. [11]For this reason, the presence of organ-specific microvasculature is essential to build physiologically relevant platforms to understand the pathophysiology of an organ.The mammary gland and breast tissue could greatly benefit from a model of mammary-specific vasculature, as breast cancer is the most commonly diagnosed cancer in women. [12]Tumoron-chip systems of various designs have been developed to study breast cancer subtypes and related processes. [13]13a] As the www.advancedscience.combreast is a dynamic system and undergoes cyclic remodeling regulated by sex hormones during puberty, lactation, menopause, and even during the monthly menstrual cycle, we expect that its microvasculature may also be sensitive to hormones.Estrogen (E2) and progesterone (P4) are steroid hormones produced and released by the ovaries.These hormones direct the development of female sexual characteristics during puberty and ensure fertility.Once E2 and P4 bind their nuclear receptors in cells, they activate the transcription of a variety of target genes involved in proliferation, metabolism, cell signaling, and survival. [14]This process leads to dynamic remodeling of the mammary gland which undergoes cycles of expansion and regression associated with changes in cell number, tissue composition, and architecture. [15]he effect of E2 on the vascular system is multifactorial, as it modulates vascular function by stimulating vasodilation and vascular relaxation by stimulating the release of vasodilatory substances from the endothelium (prostacyclin and nitric oxide synthesis) as well as by decreasing the production of vasoconstrictor agents (cyclooxygenase-derived products, reactive oxygen species, angiotensin II, and endothelin-1 (ET-1)). [16]Less is known about the effect of P4 on the circulatory system, which can have both vasoconstrictor and vasodilator effects depending on the location of the vessels and their level of exposure. [17]For example, it has been reported (in rats) that progesterone has a vasodilator effect on coronary arteries. [18]While in other studies, it is described as a vasoconstrictor during pregnancy (in rats). [19]evertheless, little is known about the effect of these hormones on organ-specific vasculature in humans since most studies use nonspecific endothelial cells in 2D systems. [20]ecognizing the importance of vascular heterogeneity, we developed a mammary-specific microvasculature in a macroscale fluidic platform wherein we can control perfusion.This model is used to investigate the role of sex hormones in mammary vascular development and barrier function.Tissue-specific mammary microvessels, generated herein, demonstrate that morphologic remodeling is promoted by fibroblast density and interstitial flow.For the first time, the role of sex hormones in the developing breast microvasculature is shown by exogenous stimulation of E2 and P4, representing the four phases of the menstrual cycle.These results demonstrate mammary microvasculature as a highly sensitive network that remodels in response to changes in fibroblast density, flow, and hormones, which are all critical in the context of the pathophysiology of breast cancer.

Interstitial Flow Promotes Mammary Microvascular Connectivity and Perfusion
Primary human mammary vascular endothelial cells (HMVECs) and human mammary fibroblasts (HMFs), sourced commercially, were first characterized by flow cytometry to confirm the expected expression of known endothelial (CD31, VE-Cadherin, endoglin) and fibroblast-associated markers (fibroblast specific protein-1, FSP1, and  smooth muscle actin, SMA), respectively (Figure S1A,B, Supporting Information).Next, HMVECs and HMFs were cocultured in a macrofluidic design previously reported and fabricated in-house. [21]Adapting our previously pub-lished protocol, [21a] cells were seeded in a fibrin hydrogel, with an endothelial to stromal cell ratio of 5:1, and after several days in culture formed a vascular network with HMFs strongly associated with vessels (Figure 1A).Although these initial experiments demonstrated the formation of microvessels, they were quite narrow and did not present perfusable vascular lumens (Figure 1A,Bii).
13a,22] Hence, we employed 3Dprinted fluidic reservoirs developed in our lab, [23] to establish an intermittent interstitial flow from day 1 onward.A computational fluid dynamics simulation demonstrates flow distributions through the device (Figure S2, Supporting Information).A range of pressure gradients (ΔP = 0-7 mm H 2 O) were examined and re-established daily.This applied interstitial flow resulted in a significant increase in perfusable vessels, ≈85%, as demonstrated by dextran perfusion, and compared to the static condition (0 mm H 2 O) with only ≈28% perfusable vessels cultured with the same volume of media (Figure 1Biii).Increasing the pressure gradient (from 3, 5 and to 7 mm H 2 O) resulted in a significant increase in vessel diameter and changes in other morphologic characteristics.However, the most physiologic vessels were generated using ΔP of 3 mm H 2 O (Figure S3, Supporting Information).
Under this pressure gradient (3 mm H 2 O used for subsequent studies), mammary microvessels form rapidly-with branched and connected vascular networks formed in less than 24 h from initial seeding (Figure 1C).21b,24] Notably, although these mammary microvessels form rapidly, they begin to regress after 4 days in culture.

Mammary Fibroblasts Affect Microvascular Morphology
21b,24,25] Fibroblasts in the breast are known to exist in lobular and interlobular positions in the mammary gland; however, little information exists about their role in vascular development and maintenance. [26]Thus, HMVECs and HMFs were cocultured at different cell ratios 5:1,10:1, and 20:1, while keeping the final concentration of endothelial cells constant (6 × 10 6 cells mL −1 ) (Figure 2A).By varying the total number of fibroblasts, a striking change in vascular morphology was observed (Figure 2A-D).An increased fibroblast concentration led to a significant decrease in microvessel diameter.Microvessels cocultured at the 5:1 ratio resulted in a mean vessel diameter of 30 ± 20 μm, comparable to in vivo values for adult women (15-50 μm). [27]The mean branch length of mammary vessels remained unchanged with varied fibroblast concentrations, yet branch density was significantly increased with increasing fibroblasts (5:1 ratio) (Figure 2C,D).Interestingly, the number of fibroblasts did not have any measurable effect on vascular barrier function (permeability to 70 kDa dextran), which was comparable among the different ratios (Figure 2E).
We established that fibroblasts were necessary to form and maintain a functional mammary microvasculature.Without the fibroblasts in coculture, thin yet connected vascular networks arise but regress rapidly after a few days in culture (Figure S1C, Supporting Information).Fibroblasts are known to secrete crucial factors like vascular endothelial growth factor (VEGF), transforming growth factor- (TGF-), and platelet-derived growth fac-tor, that promote vessel growth. [28]Therefore, we performed a semiquantitative analysis of angiogenic factors assessed for both HMVEC and HMF in monocultures (Figure 2F).Surprisingly, known angiogenic factors are expressed at relatively low levels in HMF (VEGF, basic fibroblast growth factor (bFGF), and TGF-), but inflammatory cytokines (interleukins IL-6 and IL-8, monocyte chemoattractant protein-1 (MCP-1) and epithelial-neutrophil activating peptide (ENA-78)) were expressed at relatively high levels.Moreover, we also measured MMP1, an interstitial collagenase capable of degrading collagen types I, II, and III, which is also involved in vascular remodeling. [29]MMP1 was measured from supernatant collected from the 3D microvessels grown using the 5:1 ratio at different time points.MMP1 production significantly decreases over time supporting the strong remodeling observed on the first day following seeding (Figure 2G).

Sex Hormones Alter Mammary Microvessel Morphology
The effects of 17-estradiol and progesterone were investigated on mammary vascular development.First, the expression of E2 and P4 receptors was confirmed in both HMVEC and HMF by immunofluorescence (Figure S4A, Supporting Information).Next, we generated mammary microvessels and supplemented normal media by daily perfusion of hormones at increasing concentrations (0, 1, 10, and 100 nm for E2, and 0, 1, 5, and 25 nm for P4) (Figure S4B, Supporting Information).A dose-dependent effect was observed on vascular development with high concentrations leading to nonperfusable vessels.Since E2 and P4 act in unison and are reported at much lower plasma concentrations in vivo, [30] we next examined the influence of these hormones combined as they vary across the menstrual cycle.Concentrations were chosen to mimic the mean concentrations of phases of the menstrual cycle: follicular, ovulation, luteal phase, and period (Figure 4).Hormone treatments were exogenously added to the media the day after seeding (maintaining an intermittent interstitial flow, as before) and refreshed daily for 4 days.Treatment resulted in distinct morphologic changes, particularly when high levels of either estrogen or progesterone are present, as in the ovulation and luteal phases, respectively.This is demonstrated by CD31 staining of the microvessels (Figure 3A).
To quantify vessel segments with connected and open lumens, FITC dextran was perfused in the mammary microvessels.

www.advancedscience.com
Morphologic parameters were quantified on day 4 using ImageJ, as described in our previous work. [1,24]Considering that variability between biological repeats of mammary microvessels was observed, we compared hormone-treated samples to the nontreated (normalized) control for each experiment (Figure 3C-G, with raw data in Figure S5, Supporting Information).Both the follicular and ovulation hormone treatments resulted in significant changes in the vascular area, with the ovulation concentrations leading to an increase in the vessel area coverage.The high E2 and low P4 associated with the ovulation phase also resulted in significant increases in vessel diameter (Figure 3D).These treatments led to significant changes in branch and junction densities (Figure 3E,F).All mammary microvessels treated by the luteal concentrations (highest P4) demonstrated disconnected vasculature, as shown after 4 days and thus there were no perfused vessel segments to measure with our automated segmentation pipeline (Figure 3B).
Considering that E2 and P4 can both lead to cell proliferation in certain tissues, [31] we treated HMVEC and HMF (in 2D) with concentrations representing the different menstrual phases.After four days, HMVEC endothelial cells did not respond to treatments, unlike HUVEC for comparison, as shown by EDU assays (Figure S6, Supporting Information).However, HMF proliferation was significantly increased compared to nontreated cells.

Sex Hormones Alter Mammary Microvessel Barrier Function
In addition to morphological changes observed during vessel formation, we evaluated vascular endothelial barrier function in response to sex hormones.We performed time-lapse measurements to track the flux of 70 kDa dextran across the endothelial barrier-as a proxy for endothelial permeability to solutes similar in size (such as albumin) (Figure 4A).Since the luteal phase was not perfusable, measurements were evaluated only for the other defined phases.The solute permeability of these mammary vessels was comparable across treated and nontreated conditions.Although statistically insignificant, there was an increased trend in leakiness in the ovulation phase, which had the highest amount of 17-estradiol (Figure 4B).The effect of hormones on cytokine production in these microvessels was also assessed by cytokine array (Figure S7, Supporting Information).Pooled supernatants were collected from 5 devices treated with the different hormonal conditions on day 4.Although semiquantitative, results indicate a change across the different phases in proangiogenic factors including VEGF, placental growth factor (PLGF), bFGF, and epidermal growth factor (EGF).For this reason, we further investigated these proangiogenic factors by ELISA (Figure 4C-F).Supernatant from 4 devices for the different hormonal conditions was collected on day 2 (N = 3 biological repeats), prior to any vascular regression.VEGF, which is supplemented in the media, was depleted equally between the different conditions.EGF and FGF, which are also present in the media, were differentially depleted between conditions.Not surprisingly, the follicular phase, with increased vascular remodeling, resulted in the most significant depletion of these factors.Notably, PLGF, which is not present in the culture media, is expressed at high levels in the nontreated vessels as opposed to hormone-treated conditions.It was also confirmed by ELISA that increasing lev-els of E2 results in a reduction in ET-1 (Figure S8, Supporting Information), as expected. [16]

The Effect of Sex Hormones Is Endothelial Cell-Specific
Considering the observed hormonal sensitivity of mammary microvessels, we explored the effects of E2 and P4 on HUVEC (nonmammary specific) microvessels.We cocultured HUVEC and mammary fibroblasts using the same protocol as for the mammary-specific vessels (Figure 5A); however, hormonal treatment for HUVEC-HMF cocultures lasted for 7 days, since the vessels form between days 4 and 5. Nontreated vessels resulted in thin and nonperfusable networks (Figure 5Ai).The presence of hormones drastically changes the morphology of these vessels, with a clear difference in area coverage and vessel diameter (Figure S9, Supporting Information).The presence of hormones promoted the generation of perfusable vessels.However, higher estrogen concentrations, as in the follicular and the ovular phases, resulted in very leaky vasculature, preventing permeability analysis.Expression of estrogen receptors (ER) and progesterone receptor (PGR) were examined by qPCR for both endothelial cell types and mammary fibroblasts (Figure 5B).HU-VECs highly express ER2 compared to mammary endothelial cells, and surprisingly HMFs express high levels of all hormone receptor genes.Morphologic parameters of HUVEC cocultured vessels were normalized to NT samples and compared to normalized HMVEC cocultured vessels (Figure 5C-F).The comparison shows a clear difference in the relative changes in diameter and area between the two cocultured microvessels as they are treated with hormones.

Discussion
The generation of blood vessels in vitro is crucial for mimicking in vivo organ and tissue function, as blood vessels are essential for characteristic tissue growth, properties, and maintenance.Several organ-on-chip models have successfully integrated a vascular network, with remarkable success in the lung, liver, skin, blood-brain barrier, and kidney. [33]One of the limitations of vascularized organ-on-chip technology is the use of HU-VEC and other nonspecific endothelial cells. [34]13a] Little is known about the vasculature in the dynamic mammary gland, thus, for the first time, perfusable mammary microvascular networks with breastspecific cells were generated.
Herein, HMVEC-HMF cocultures were optimized to generate perfusable mammary microvessels.These vessels demonstrate rapid vascularization, characterized by the appearance of an extensively branched and connected endothelial network as early as 24 h postseeding, with full perfusion by days 3-4.21b,24] In the mammary system, accumulation of cells in the vessel lumen and subsequent negative remodeling of vessels occurs by day 5 (Figure 1C) limiting long-term in vitro studies.Our cocultures are generated from the only current commercial source of human mammary endothelial cells and fibroblasts.Although generating mammary microvessels from other donors would be of significant interest, isolating breast endothelial cells from reduction mammaplasties has a very low success rate and they do not typically survive in culture. [35]lood flow is necessary for the delivery of oxygen, immune cells, and nutrient delivery.Changes in vascular dynamics are well-known to affect the behavior of endothelial cells, which has been recapitulated in vitro. [36]Here, gravity-driven flow was established in our microvessels through a 3D printed reservoir, as interstitial flow has been previously shown to induce early vessel formation. [37]Mammary microvessels in our system are highly mechanosensitive, with higher pressure gradients resulting in significant changes in morphology and barrier function (Figure S3, Supporting Information).27b] Heterogeneity in branch patterns occurred near regions of increased interstitial shear stress, as shown by a computational fluid dynamics (CFD) simulation of the device (Figure S2, Supporting Information).Mammary microvessels demonstrate similar behavior to HUVEC vessels, where an applied interstitial flow was shown to promote vessel formation, diameter enlargement, and reduced permeability to solutes. [36]2c,10] Previous work from our group has shown that tissue-specific stromal cells modulate vascular morphology and barrier function, highlighting the need for the use of organspecific cells. [24]Moreover, fibroblasts have been shown to necessarily support lumen formation and longevity of microvessels in vitro through transcriptomic analysis. [38]Therefore, we hypothesized that mammary fibroblasts are important contributors to mammary vessel formation.Notably, mammary endothelial cells cultured alone show discontinuous vasculature in the absence of fibroblasts (Figure S1C, Supporting Information).The presence of fibroblasts allows the formation of vasculature and there is a strong correlation between fibroblast density and vascular morphology.27b] MMP1, a metalloprotease that degrades collagen type 1, is strongly expressed in the first days of coculture, indicating that matrix degradation occurs during the earliest stages of network formation.The mammary gland is rich in stromal tissue and fibroblasts are the predominant stromal cell type.Under physiological conditions, fibroblasts are essential in the mammary gland and change during the different hormonal phases. [39]During pregnancy and lactation, for example, fibroblasts regulate the composition, organization, and stiffness of the extracellular matrix (ECM), which controls epithelial cell proliferation, migration, differentiation, and polarity, facilitating the formation of new lactiferous alveoli and supporting blood microvessels. [39]Fibroblast function is strongly influenced by sex hormones herein, as shown by the fact that mammary fibroblasts express high gene expression levels of estrogen and progesterone receptors in comparison to endothelial cells (Figure 5B).Nevertheless, the effect of sex hormones on breast fibroblasts remain poorly understood, as most studies on these cells have focused on their role in breast cancer, in which, cancer-associated fibroblasts (CAFs) are the most studied.CAFs represent the predominant cell population (≈80%) in the tumor microenvironment and are involved in breast cancer development, progression, and metastasis. [40]Investigating contributions of breast fibroblasts in physiologic states will enhance our understanding of their role in regulating breast microvasculature in health and disease.
Hormone oscillations during the menstrual cycle affect the mammary epithelium, ECM, stromal compartment, immune system, and vascular network. [41]Previous 2D studies on endothelial cells reported an increase in proliferation and angiogenesis in the presence of estrogen. [42]Additionally, alterations in vascular density within the mammary gland have been documented during the phases of pregnancy and lactation, suggesting a direct influence of sex hormone concentration on vascular growth and remodeling. [43]In order to shed light on the effects of E2 and P4 on vascularization and endothelial barrier function, we perfused our mammary-specific microvessels with hormone concentrations corresponding to the different phases of the menstrual cycle. [44]Although hormone concentrations fluctuate during the menstrual cycle, for practical reasons (a practical limitation here) we chose an average value representing each phase.After 4 days of hormone treatment, striking changes in vascular morphology were observed (Figure 3A).In particular, hormones that mimic the luteal phase (high P4) resulted in progressive regression of vessels.One explanation for a reduction in vascular density could be due to enhanced proliferation and competition of HMFs, which are proliferative in response to these treatments in 2D (Figure S6C, Supporting Information).Significant vascular remodeling was also observed in the follicular phase-decreased vascular area and diameter and increased branch and junction densities.This result could be explained by E2 peaks in the follicular and luteal phases, which are recognized as highly mitotic phases in the breast. [45]In fact, it is known that with increasing estrogen and progesterone, epithelial alveolar buds start to form and differentiate. [41,46]In the ovulation phase (highest E2 concentration) significantly larger vessel areas and diameters were observed with reduced branch density.Despite the variability between experiments, a trend toward increased leakiness of the vessels can be observed during the ovulation phase (Figure 4B).16d,47] As further evidence, a vasoconstrictor like ET-1, which has been demonstrated to decrease in the presence of estrogen in HUVECs, also exhibits downregulation in our system across different phases of the menstrual cycle (Figure S8, Supporting Information). [48]Analysis of four (VEGF, FGF, EGF, and PLGF) important proangiogenic factors were also measured in response to hormone treatments.VEGF, which plays a key role in angiogenesis and is present in the vas-cular media, was depleted in all conditions (hormone treated or not) indicating its importance in the formation and maintenance of the microvasculature.Of note, concentrations of PLGF, which are not present in the cell culture media, showed a similar trend to FGF and EGF, with depletion in the follicular phase.PLGF belongs to the VEGF family and promotes vascularization, therefore it is possible that in the follicular and ovulatory phases, with increased vascular remodeling, there is a negative feedback mechanism leading to reduced concentrations in comparison to untreated microvessels.Of course, the breast is much more complex than our reductionist model system, and other cell types (adipose, immune, epithelial and smooth muscle cells) are also influenced by hormones, which affect the microvasculature.Additionally, other hormones such as follicle-stimulating hormones and luteinizing hormones also have an effect on the breast, this has not yet been fully elucidated and could be investigated in the future using our model.Although beyond the scope herein, the intricate interactions between various cell types and hormones can be effectively explored through transcriptomic characterization, as was recently performed by Murrow et al. with breast tissue.The researchers employed single-cell analysis to map breast cell response to hormonal oscillations across individuals, underlying that events like pregnancy and obesity could change the hormone-responsiveness in the breast. [49]We observed a clear effect of sex hormones on mammary microvessel in our model, and also showed that hormone-responsiveness was not unique to HMVEC, as HUVEC-HMF cocultures were also influenced by hormone treatments.Differences between the two endothelial cultures treated are notable; particularly in the luteal phase where mammary-specific vessels led to vessel regression after 4 days, yet treatment in HUVEC microvessels results in a fully functional vascular network (Figure 5A).One potential explanation for the observed difference is that HUVECs typically are exposed to high levels of P4 nearing gestational term (although ≈1% of the total in maternal serum), [50] so it is possible this reinforces proangiogenic behavior with these cells.
Vessel diameter increases with all hormone treatments in HU-VEC microvessels (Figure S9, Supporting Information), while in the mammary microvessels, diameter only increased significantly between the follicular and estrous phase (Figure S5, Supporting Information).One explanation for these cell-specific differences is likely due to different expression levels of ER and PR in HUVEC and HMVEC (Figure 5B).Moreover, HMVEC cells originate from a single female donor, whereas HUVEC are from pooled (male and female) donors.Pooled donors reduce donor-to-donor variability; however, experiments using endothelial cells from different donors should be a focus of future works.Our research underscores the significance of utilizing tissuespecific cells to investigate organ-specific characteristics, such as the function of sex hormones in breast microvasculature.

Conclusion
For the first time, perfusable primary mammary microvessels have been cultured on-chip and used to demonstrate the effect of physiological concentrations of sex hormones.Estrogen and progesterone are shown to have a profound effect on breast-specific and nonspecific microvessels at menstrual level concentrations.Mammary fibroblasts play a prominent role in microvascular development and also appear to contribute to the hormone-responsiveness of microvessels.This work highlights the importance of using tissue-specific cells and the role of sex hormones in breast vascular remodeling.In the future, this system will enable the study of pregnancy and menopause hormone levels in breast microvessels.In addition, this system could be used as a starting point for building a vascularized breast tumor on-chip.The presence of mammary-specific vasculature will provide insightful information about the drug delivery process and its impact on the healthy and diseased mammary microenvironment.

Experimental Section
Cell Culture: Commercially available HMVECs were purchased from Innoprot (Cat # P10892) and were cultivated in endothelial media (Vas-cuLife, Lifeline cell systems) with 5% fetal bovine serum (FBS) on 30 μg mL −1 human fibronectin (Sigma) coated T-75 flasks.Primary human mammary fibroblasts were purchased from Innoprot (Cat # P10893) and were cultured in Fibrolife media (Lifeline cell systems) on 50 μg mL −1 rat tail collagen I (Merck) coated T-75 flasks.All cells were used between passages 3 and 5. HUVECs were purchased from Lonza and transduced with viral particles generated in the lab by standard lentiviral particles production using the 3rd generation lentiviral expression vector pLV-mCherry (pLV-mCherry was a gift from Pantelis Tsoulfas) and cultured in endothelial media VascuLife on 50 μg mL −1 rat tail collagen I coated flasks.mCherry-HUVECs were used between passages 6 and 9.For all the cell lines used, the dissociations were carried out using TrypLE Express (Gibco), and growth media was completely refreshed every other day in the 2D cultures.
Device Fabrication: As previously published, [21c] devices were fabricated using PDMS (SYLGARD 184 Silicone Elastomer Kit, Dow).The elastomer and cross-linker were mixed in a 10:1 ratio as per the manufacturer's recommendation.Once mixed the PDMS was poured in a prefabricated multidevice negative mold and degassed using a vacuum desiccator.PDMS was then cured overnight at 60 °C and then each device was cut and punched.Each PDMS device was air-plasma bonded (Harrick systems) to clean glass slides.The devices were coated with 30 μg mL −1 human fibronectin (Sigma) for 1 h at 37 °C and after that the devices were placed in the 60 °C oven for 24 h to return to its native hydrophobic state.
Microvasculature Formation: Fibrinogen derived from bovine plasma (Sigma) was reconstituted in phosphate-buffered saline (PBS) to a working concentration of 6 mg mL −1 before use.Thrombin (Sigma) was diluted to a 4 U mL −1 working solution in cold VascuLife medium.HMVECs were stained with CellTracker Green CMFDA Dye (ThermoFisher) before the seeding to visualize vessel structures.The HMVEC and HMF were cultivated for two passages and when the confluence was reached they were dissociated and resuspended at 12 million cells per mL and 2.4 million cells per mL (5:1 ratio) in 4U thrombin.As previously described, [21c] the cells in thrombin were mixed with 50% v/v of fibrinogen to make up final concentrations of 6 million cells per mL and 1.2 million cells per mL in 3 mg mL −1 fibrin gel.The gel-cells mixture is inserted in the central channel of the device and to allow the polymerization of fibrin the devices were left at 37 °C for 10-20 min.VascuLife with 5% FBS media was used to fill the media channels.After 1 day from the seeding, 3D printed custommade reservoirs were inserted and used to generate an intermittent interstitial flow.For the 3 mm H 2 O pressure the media in the reservoir chamber is respectively 600 and 240 μL, while for the condition 0 mm H 2 O pressure in both reservoir chambers the media volume is 420 μL, refreshed daily.Microvessels were grown over 4 days prior to any treatments or permeability measurements.
21b,24] A 0.1 mg mL −1 solution of 70 kDa TexasRed labeled dextran (ThermoFisher) in complete growth media (Vasculife) was perfused through the microvessels by a generated pressure drop across the gel.Both media channels were aspirated and then 40 μL of dextran solution was first added to one media channel, to allow flow across the gel.This was then stabilized with the addition of an equal volume dextran solution to the opposing media channel to stop convective flow (≈30 s later.).All images were acquired on a Stellaris 8 confocal microscope using LAS X software (Leica).Intervals were set to 3 min between z-stacks.After 1 min, time-lapse (3 × 3 min intervals) confocal z-stack images were acquired at a 5 μm step size and ≈20-25 slices.21b] The morphological quantifications were then normalized to the size of the imaging region, as these measurements were performed on FITC and Texas red-dextran channels.
Cell Extraction from Microfluidic Devices: The cell-gel channel of the devices was cut using a safety razor blade (VWR, Cat # 700-0418).The gels were digested in a solution of Accutase (Gibco) and 50 FU mL −1 Nattokinase (Japan Bioscience Ltd.) for 15-20 min at 37 °C.Afterward, the content was homogenized and centrifuged.The remaining pellet was resuspended in 1 mL of RNase-free water and stored at −70 °C.
RNA Extraction and Reverse Transcription in cDNA: RNA was extracted using the RNeasy Mini Kit (Qiagen, Cat # 74134) and performed according to the manufacturer's handbook.The RNA samples were diluted using 20 μL RNase-free water.The isolated RNA was converted to complementary cDNA using the High-Capacity RNA-to-cDNA Kit (Applied Biosystems, Cat # 4387406).Depending on the targeted cDNA concentration, the RNA volume for the reverse transcriptase was determined.The cDNA synthesis was performed in a thermocycler according to the manufacturer's handbook.After reverse transcription, cDNA samples were stored at −20 °C until usage.
RT PCR: To investigate the gene expression of the estrogen and progesterone receptors TaqMan real-time PCR was performed on 50 ng of cDNA per samples.The assay was performed using TaqMan Fast Advanced Master Mix (ThermoScientific, 4444556).The following primers were used to analyze the ER1 gene (ThermoScientific, Hs01046816), ER2 gene (ThermoScientific, Hs01100353), and PGR (ThermoScientific, Hs01556702).The assay has been performed in a 96-well plate (LightCycler 480 Multiwell Plate 96, white, Roche) on the LightCycler 480 II (Roche).The PCR program includes a preincubation step at 50 °C for 2 min and 95 °C for 2 min, followed by 40 cycles of 1 s at 95 °C and 20 s at 60 °C for the amplification step.The gene expression data were normalized to the mean of two housekeeping genes ACTB (ThermoScientific, Hs01060665), TUB1A (ThermoScientific, Hs03045184), and tested in triplicates.The experimental data were analyzed with the 2-ΔCt method.
Immunofluorescence Staining: Fixation of cells and or devices was performed using 4% paraformaldehyde for 20 min prior to washing with PBS and subsequent solubilization using 0.1% Triton-X (10 minutes).To stain microvessels within a device, a pressure gradient was applied across the gel for all staining and wash steps.Samples were then incubated in applicable blocking buffer, PBS + BSA + serum of the secondary antibody, for more than 1 h for 2D samples or overnight for the staining for the microvasculature.Primary antibodies were diluted in wash buffer (0.5% BSA in PBS) and were added to the samples and incubated overnight at 4 °C.The primary antibodies used in the experiments were: Recombinant Anti-Estrogen Receptor alpha (phospho S118) antibody (Abcam ab32396, 1:100), Anti-Progesterone Receptor (phospho S190) antibody (Abcam, ab131110 1:100), Anti-CD31 antibody (Abcam, ab187377, 1:500), Recombinant Anti-S100A4 antibody (Abcam ab124805 1:500), and DAPI (FisherScientific, D3571) for the nuclei counterstaining.After overnight incubation, samples were washed with wash buffer and incubated with the appropriate secondary antibodies and counterstains (>2 h).Samples were rinsed with PBS and either imaged immediately or mounted on coverslips (for 2D samples) using Fluoromount-G (Invitrogen) and stored at 4 °C before imaging.
Cytokine Array: For 3D cytokine analysis, supernatants were collected from 2D culture of HMVEC and HMF pooled from n = 5 devices on day 4. A human angiogenesis array (Abcam, ab134000) was employed according to the manufacturer's instructions.The relative expression of cytokines (measured by chemiluminescence intensity) was compared between all groups, corrected to the negative controls on each array, and normalized to the positive controls, using the monoculture as the reference array.The blots were visualized and imaged (1 min exposure) using the Fusion FX Spectra (Vilber, France).
ELISA Assay: Quantification of ET-1, VEGF, bFGF, PLGF, and EGF in the media collected from each device (N = 3 with 3 devices per experiment) was performed respectively with ELISA assay Human Endothelin-1 Quantikine ELISA kit (R&D Systems, DET100): Human VEGF Quantikine ELISA Kit (R&D Systems, DVE00), Human FGF basic/FGF2/bFGF Quantikine Kit (R&D Systems, DFB50), Human PlGF Quantikine ELISA Kit (R&D Systems, DPG00), and Human EGF Quantikine ELISA Kit (R&D Systems, DEG00) according to manufacturer's instructions.For this experiment, the supernatant was taken from the devices for the different hormone conditions on day 2 to exclude any influence of other processes involved in vessel regression, which could occur on day 4.
MMP1 Expression: Using DuoSet ELISA KIT (R&D systems, DY901B), serum MMP-1 levels were determined for microvessels on day 1 and 4 per the manufacturer's protocol.Media samples were pooled from n = 5 devices, immediately placed on ice, and frozen down at −70 °C until assays were run.Samples were thawed on ice before the assay, while the other reagents were brought to room temperature before use (per manufacturer's instructions).An appropriate sample dilution of 1:10 was determined to fit within the range of the MMP1 standard curve, using microvessel growth media as media control.Concentrations were measured using a plate reader at 450 nm (and a wavelength correction at 590 nm).Absorbance values were compared to provide standards, accounting for the sample's dilution factor.
Proliferation Assay: To measure the proliferation of the cells in response to hormone treatment Click-iT Plus EdU Imaging Kits Protocol (Invitrogen,c10639) was used according to the manufacturer's instructions.Briefly, after 4 days of treatment with hormones, the samples were stained with EdU labeling solution (final concentration of 10 μm) for 2 h.Then the cells were fixed with 4% PFA and permeabilized with (0.5% Triton X-100 in PBS per 20 min).Later the cells were treated with Click-iT Plus reaction cocktail for 30 min and then the nuclei were stained with Hoechst 33342 for 30 min.The images (4 areas per well) were acquired with Thunder Imager Live Cell & 3D assay and using LAS X software (Leica) using the 10× objective.
Statistical Analysis: Statistical significance was analyzed using Orig-inPro v.9.85.For the samples that do not reject normality the one-way ANOVA was used to assess statistical significance across conditions at p < 0.05, and a post hoc Tukey test was performed as a means comparison, where differences at p < 0.05 were taken as significant (*, p < 0.01 **, p < 0.001 ***, p < 0.0001 ****).The samples that reject normality the Kruskal-Wallis ANOVA were used to assess statistical significance across conditions.
CFD Simulations: CFD simulations were used to approximate the flow-induced shear stresses within the gel chamber of the microfluidic device.A CAD design of the complete microfluidic device including gel chamber that was sandwiched between fluid channels (.dxf file) was imported to COMSOL Multiphysics software (Version 6.0).This served as the geometrical input for the CFD simulation.Boundary conditions included the applied pressure gradient across the gel chamber (top to bottom) of the chip as shown in Figure S2 of the Supporting Information and no-slip conditions were applied to the side walls.Inlet pressure is set at 0.3 mbar (30 Pa or 3 mm H 2 O) and outlet at 0 mbar. [52]For this study, fluid properties of cell culture media were modeled as Newtonian fluid resembling water with a density of 998.2 kg m −3 and dynamic viscosity 9.4 × 10 −4 Pa s.Gel properties were modeled as having a density of 985 kg m −3 and dynamic viscosity of 1 × 10 −2 Pa s.Moreover, the porosity and permeability of the gel used were 0.3 and 1 × 10 −13 m 2 , respectively.To predict the flowinduced shear stresses and velocities profile within the gel chamber of a microfluidic chip, the Brinkman equation physics (flow through porous media) of COMSOL was used with physics-defined extremely fine mesh conditions.

Figure 1 .
Figure 1.A) (left) Graphical representation of the procedure used to generate mammary microvessels.Representative confocal maximum projection of fixed and stained microvessels at the fourth day in culture.Microvessels stained with endothelial marker CD31 green, fibroblast-specific protein-1 (FSP-1, red) and nuclei with DAPI (blue).Scale bar is 100 μm.B) Vessels generated using a i) custom-made 3D printed reservoir to induce a hydrostatic pressure gradient, ΔP. ii) Representative confocal images of microvascular networks generated at different conditions: static culture without and with a reservoir under static conditions (ΔP = 0 mm H 2 O) and flow conditions (ΔP = 3 mm H 2 O).HMVEC in the system are visualized by CellTracker green and the vessels perfused with Texas Red 70 kDa dextran.Scale bar is 200 μm.iii) Percentage of perfused vessels in all three culture conditions for N = 2 biological repeats.Box plots demonstrate median, percentile 25-75 quartile (box edge) and 10-90 (outer whiskers).Significance is shown by *p < 0.05 and ***p < 0.001 using one-way ANOVA with Tukey means comparison test.C) Representative confocal images of microvascular network at different days in culture.HMVECs in the system are shown by CellTracker green and perfusable vessels by Texas Red 70 kDa dextran.Scale bar is 200 μm.The gray bar indicates the lack of perfusion over time of the vessels.

Figure 2 .
Figure 2. A) Confocal maximum projection images highlight microvessel formation at varied endothelial to fibroblast cell ratios.The gray bars graphically represent the number of fibroblasts and endothelial cells in the different conditions.HMVECs are shown with CellTracker green and microvessels are perfused with Texas Red 70 kDa dextran at day 4 to assess vessel perfusion.Scale bar is 200 μm.Morphologic comparison of the microvessels at the different ratios: B) effective diameter (blue region shown demonstrates average ex vivo mammary-specific measurements), C) mean length, D) branches density, and E) permeability to 70 kDa dextran all for N = 2 biological repeats.F) Semiquantitative measure using a cytokine array of supernatant collected from HMVEC and fibroblast (HMF) cells in monolayer culture.G) MMP1 quantification by ELISA of control media (VL 5%) and supernatant from devices (5:1 ratio) from day 1 to 4. Box plots demonstrate median, percentile 25-75 quartile (box edge) and 10-90 (outer whiskers).Significance is shown by *p < 0.05, **p < 0.01, ***p < 0.001, for data following normality with one-way ANOVA with Tukey means comparison test, and if normality is rejected using Kruskal-Wallis ANOVA test.

Figure 4 .
Figure 4. A) Confocal maximum projection images of mammary microvessels formed under different hormonal conditions at day 4. HMVEC shown by CellTracker green and microvessels perfused with Texas Red 70 kDa dextran at day 4 to assess vessel permeability.Scale bar is 200 μm.B) Vessel permeability to 70 kDa dextran is shown for the different hormonal conditions (nontreated, NT; Follicular, Fol; Ovulation, Ovu; Luteal, Lut).Data from five separate experiments with ≥3 devices per condition.ELISA assay performed from supernatant of devices collected at day 2 for C) VEGF, D) FGF, E) EGF, and F) PLGF.Data are from 3 biological repeats with 3 devices per condition each.The dashed red lines in the plots indicate the level of each factors present in the control media.Box plots demonstrate median, red dots for the mean, percentile 25-75 quartile (box edge) and 10-90 (outer whiskers).Significance is shown by *p < 0.05, **p < 0.01, ***p < 0.001, using one-way ANOVA with Tukey means comparison test for data following normality, or if normality is rejected using Kruskal-Wallis ANOVA test.