Sex-Specific Response to Combinations of Shear Stress and Substrate Stiffness by Endothelial Cells In Vitro

Abstract

By using a full factorial design of experiment, the combinatorial effects of biological sex, shear stress, and substrate stiffness on human umbilical vein endothelial cell (HUVEC) spreading and Yes-associated protein 1 (YAP1) activity are able to be efficiently evaluated. Within the range of shear stress (0.5–1.5 Pa) and substrate stiffness (10–100 kPa), male HUVECs are smaller than female HUVECs. Only with sufficient mechanical stimulation do they spread to a similar size. More importantly, YAP1 nuclear localization in female HUVECs is invariant to mechanical stimulation within the range of tested conditions whereas for male HUVECs it increases nonlinearly with increasing shear stress and substrate stiffness. The sex-specific response of HUVECs to combinations of shear stress and substrate stiffness reinforces the need to include sex as a biological variable and multiple mechanical stimuli in experiments, informs the design of precision biomaterials, and offers insight for understanding cardiovascular disease sexual dimorphisms. Moreover, here it is illustrated that different complex mechanical microenvironments can lead to sex-specific phenotypes and sex invariant phenotypes in cultured endothelial cells.

1. Introduction

In recent years, biomaterials research has recognized the importance of not only chemical factors, but also mechanical factors for directing cell fate.^[1–3] This realization has paved the way for organ-on-a-chip models, tissue models, and tissue engineered systems that aim to better mimic the in vivo environment in vitro.^[4–8] In pursuit of this, tremendous efforts have been made to realize the role of cell–material interactions on cell behavior from material elasticity and viscoelasticity to cell memory of their past environments.^[9–17] In the same vein, the influence of physical forces such as fluid shear stress from blood flow and interstitial flow and bulk tissue deformations have been recognized for their role to mechanically regulate cell response.^[18,19] But rarely have we as a field investigated the combination of mechanical stimuli that can interact to be more than the sum of their parts.^[20–24] For instance, in cultured endothelial cells, the combination of varying levels of shear stress and substrate stiffness synergistically and antagonistically mediated cell elongation, alignment, spreading, mechanics, and inflammatory and vasoregulatory behavior.^[22,23] Though these studies illuminated much about the interaction between shear stress and substrate stiffness, sex was not included as a biological variable.

This is unsurprising as sex as a biological variable has been absent from most in vitro work.^[25–28] In 2011, it was highlighted that cell culture models have ignored sex as an experimental variable–cell sex went unreported in 72% to 80% of published articles in leading cardiovascular-specific journals.^[26] In the ten years since, reporting has improved, but still remains inconsistent and below 50% of published articles in the cardiovascular field.^[27] In 2017, it was reported that 62% of nonreproductive organ-on-a-chip technologies did not specify sex and for those that did specify sex, the cells used were predominantly male. Most recently, we have shown in a review of the biomaterials literature from December 2019, that 96% of bio-material studies failed to report the sex of their cells.^[25] This omission reiterates that sex as a biological variable largely has not been considered in the field.

Despite this, there is significant reason to include biological sex in biomaterials research.^[25] It has been shown that osteoblasts display sex-specific and surface-dependent responses to systemic hormones while investigating different titanium surface topographies for orthopedic implants.^[29,30] Materials in another form, as either nano- or micro- particles have been shown to elicit sexually dimorphic responses in terms of blood–brain barrier crossing following injury, gold nanoparticle toxicity, and wear particle sensitivity.^[31–34] Relevant to tissue engineering strategies, human pluripotent stem cells display sex-specific differences in differentiation efficiency indicating that protocols should be tailored to cell sex.^[35] Together these results begin to reveal among others that to provide precision medicine and precision biomaterials will require expanding our knowledge of sex-specific cell–material interactions.^[7,8,36] How can we create systems that are robust to many different patients aimed at providing precision medicine, if the systems have not been challenged to recapitulate the features inherent to differences in biological sex? Because the field is so open—little research exists at the intersection of biomaterials and sex—we need a map to help direct where we should focus our energy and resources.

This exploratory and observational study intends to be the basis for that map, to champion the importance of sex as a biological variable in the biomaterials field. For this we turned to the well characterized and mechanobiologically active cell type, the endothelial cell. We hypothesized that endothelial cells would gain and lose sexually dimorphic behavior in response to different mechanical microenvironments. This is based on the strong evidence that mechanical stimuli interact to regulate endothelial cell phenotype and that endothelial cell phenotype is sexually dimorphic in static culture. Toward this aim, we conducted a 2³ full factorial study to define regions of interest at the intersection of biological sex and mechanical stimulation. We cultured male and female human umbilical vein endothelial cells (HUVECS) as a model endothelial cell type and stimulated them by physiological combinations of fluid shear stress and substrate stiffness relevant to the vascular niche. We then measured for cell morphometrics and mechanosensitivity metrics as simple, informative indicators of sex-specific response. Herein, we provide a revealing assessment of the male and female endothelial cell response to combinations of shear stress and substrate stiffness.

2. Design of Experiment

A full factorial design is an efficient strategy for exploring a multi-factor space. We used a 2^k full factorial design to assess both main and interactive effects of three factors, sex, shear stress, and substrate stiffness on endothelial cell response. For our design we used two levels (high and low) for the corner points to be able to fit a linear model to our measured response metrics as a function of our three factors of interest. However, this design alone would not inform us about the response between the levels, i.e., if the response displayed any nonlinear behavior. Thus, we modified the design with center points to indicate any curvature in the response without drastically increasing the number of experimental runs.^[51] For a two-level design the center points are simply the overall midpoint between the high and low levels across all continuous factors. Additionally, a factorial design is the suggested method according to the National Institutes of Health Office of Research on Women’s Health for investigating the effect of sex and other factors because it provides a way to identify interactive effects.^[52,53] The levels for each continuous factor (shear stress and substrate stiffness) were chosen based on literature values for the ranges experienced in the vasculature. For shear stress, the low corner point was set at 0.5 Pa (5 dynes cm⁻²). This amount of shear is considered to be atheroprone. The high corner point was set at 1.5 Pa (15 dynes cm⁻²) as this amount of shear stress promotes realignment, is considered atheroprotective, and is the average wall shear stress in arteries.^[1] As for substrate stiffness, the low and high corner points were set at 10 and 100 kPa because this range spans that of the endothelial cell basement membrane.^[1] Thus, the center points were set as 1.0 Pa (10 dynes cm⁻²) and 55 kPa for both males and females (Figure 1). A complete set of runs included 10 factor-level combinations (Table 1). Three independent experimental replicates were conducted at each factor-level combination using populations of pooled male and pooled female endothelial cells each derived from three donor populations.

Figure 1.

Visual representation of the experimental design. Corner points in black. Center points in red.

Table 1.

2³ Full factorial plus center point design of experiment for one replicate.

Run order	Sex	Shear stress [Pa]	Stiffness [kPa]	Point
1	Male	0.5	10	Corner
2	Male	1.5	10	Corner
3	Male	0.5	100	Corner
4	Male	1.5	100	Corner
5	Male	1.0	55	Center
6	Female	0.5	10	Corner
7	Female	1.5	10	Corner
8	Female	0.5	100	Corner
9	Female	1.5	100	Corner
10	Female	1.0	55	Center

The value of using a 2^k full factorial design is that it simplifies the data analysis by being able to fit the data to a linear model. In this case, a saturated linear model with center points was used (Equation (1)) where Y is the response, the β terms are model coefficients, X are the factors, and ϵ is the model error

$$Y={\beta }_0+{\beta }_1{X}_1+{\beta }_2{X}_2+{\beta }_3{X}_3+{\beta }_{12}{X}_1{X}_2+{\beta }_{13}{X}_1{X}_3+{\beta }_{23}{X}_2{X}_3+{\beta }_{123}{X}_1{X}_2{X}_3+{\beta }_{Ct}{X}_{Ct}+ϵ$$ (1)

Fitting to the model enables evaluation of statistically significant effects (X), indication of nonlinearity in the response, and a degree of prediction within the range of the corner points.

3. Results

3.1. Sex Validation

Polymerase chain reaction (PCR) analysis confirmed that the cell sex as reported on the certificates of analysis from the vendor was accurate for each population of HUVECs. It is necessary to make this validation as this information may not be provided or be accurate from the vendor.^[25] PCR was used to amplify in a single reaction a region of the sex-determining region Y (SRY) gene, specifically the region for testis-determining factor (TDF) (males only), and a region of the HLA-DQα gene, an internal control found on chromosome 6 (both males and females).^[54] (The protocol uses two primer sets (Table 2) to amplify in a single reaction a region of the SRY gene, specifically the region for TDF (males only), and a region of the HLA-DQα gene, an internal control found on chromosome 6 (both males and females)). Thus, sex determination could be assessed as males would have two PCR products and females would have one PCR product. DNA gel electrophoresis of the PCR products revealed the sex of each cell population. In the gel, all three male populations displayed two bands, one at 139 bp (TDF, common only to males) and a second doublet band at 239/242 bp (HLA-DQα, common to both males and females). Conversely, all three female populations displayed one doublet band of PCR products for HLA-DQα (Figure S1, Supporting Information). Following this confirmation, the mechanical stimulation experiments were conducted according to the prescribed combinations (Table 1).

Table 2.

Primer sequences used for sex confirmation.

Primer	Direction	Sequence
TDF	Forward	5′ ATGACCCTAGAACCACTGGA 3′
TDF	Reverse	5′ GAGTATTGCGTTGGCATCCT 3′
DQα	Forward	5′ GTGCTGCAGGTGTAAACTTGTACCAG 3′
DQα	Reverse	5′ CACGGATCCGGTAGCAGCGGTAGAGTTG 3′

3.2. Morphometrics and Mechanosensitivity Metrics

For our exploratory study, we evaluated the cell response by informative metrics of morphology and mechanosensitivity, namely, the morphometrics, cell area and nuclear area, and the mechanosensitivity metric, the ratio of nuclear Yes-associated protein 1 (YAP1) to cytoplasmic YAP1. These metrics have proved to be increasingly informative as they have recently been shown to inform cell phenotype classification when using neural network schema broadly available MATLAB.^[55] We evaluated these two morphometrics because they are sensitive to each individual factor, sex, shear stress, substrate stiffness.^[1,46] Similarly, we measured YAP1 localization for its sensitivity to different mechanical stimuli and different levels of those stimuli. Mechanical stimulation regulates YAP1 localization to either the cytoplasm (inactive) or the nucleus (active). For instance, its localization has been correlated to atherogenic and atheroprotective endothelial cell phenotypes after stimulation by different types of flow profiles.^[56–58] Measurements were made on a single cell basis for male and female HUVECs immunostained for their cytoskeletal F-actin (Figure 2) and their YAP1 content (Figure 3) following mechanical stimulation.

Figure 2.

Representative fluorescent images of cytoskeletal F-actin staining at each factor-level combination. Scale bar 50 μm.

Figure 3.

Representative confocal fluorescent images of YAP1 immunostaining at each factor-level combination. Scale bar 50 μm.

In general, both sexes progressed from a circular shape with random organization to a more elongated shape with alignment in the direction of flow as both shear stress and substrate stiffness increased (Figures 2 and 3). Overall, there appeared to be a qualitative enhancement to elongation and alignment for the female HUVECs compared to the male HUVECs. At each factor-level combination, more than 700 cells were evaluated for their morphometrics and mechanosensitivity. All three metrics, cell area, nuclear area, and YAP1 nuclear-to-cytoplasm ratio followed log-normal distributions at all factor-level combinations (Figure 4). Details of the descriptive statistics (number of cells measured, mean, standard deviation, interquartile, and median) for each metric at each factor-level combination are available in the Supporting Information (Tables S1–3, Supporting Information).

Figure 4.

Violin plots for each factor-level combination of A) cell area, B) nuclear area, and C) YAP1 nuclear-to-cytoplasm ratio. The large dashed lines indicate the median value and the small dashed lines indicate the 25th and 75th quartile values. Plots for male HUVECs are shown in blue (left of pairs) and plots for female HUVECs are shown in red (right of pairs).

To visualize the male and female HUVEC response to the different mechanical microenvironments of shear stress and substrate stiffness, the median value of the metrics was plotted as linear interpolated response surfaces (Figure 5) Male HUVEC cell area showed a pyramidal response surface with a maximal value at the combination of high shear stress and high substrate stiffness (Figure 5A). Conversely, female HUVEC cell area showed a fairly planar response with its maximal value at the combination of high shear stress and low substrate stiffness (Figure 5B). Moreover, from the combined plot of male and female HUVEC response it became apparent that female HUVECs were larger than male HUVECs for most stimulation combinations. Nuclear area responded differently. Male HUVEC nuclear area took on a skewed saddlehorse shape being more elevated toward the higher stimuli conditions whereas female HUVEC nuclear area featured a large minimum at the center point (Figure 5D–F). The shape of the response surfaces for YAP1 localization ratio mirrored those for cell area (Figure 5G–I). Immediately, it became clear though that YAP1 nuclear content for female HUVECs was largely invariant to any of the stimuli combinations. The median ratio remained between 1.53 and 1.61, changing only slightly within that narrow range across all factor-level combinations. The median value trended linearly with minimal slope within the range of shear stress and substrate stiffness combinations. In contrast the median ratio for male HUVECs ranged from 1.21 to 1.58 and displayed nonlinear, quadratic behavior. Only under conditions of the center point and the high shear stress and high substrate stiffness corner point did the YAP1 nuclear-to-cytoplasmic ratio for male HUVECs approach the value maintained by female HUVECs across all stimulation conditions. Overall, female HUVEC YAP1 localization was robust to changing stimuli, i.e., it was largely invariant to mechanical stimuli while male HUVEC YAP1 localization was much more sensitive. From the surface plots it became clear that the overall trends in the response between male and female HUVECs to the range of shear stress and substrate stiffness were sex-specific.

Figure 5.

Surface plots of the median A) cell area, D) nuclear area, and G) YAP1 nuclear-to-cytoplasm ratio for male HUVECs at each factor-level combination of shear stress and substrate stiffness. Surface plots of the median B) cell area, E) nuclear area, and H) YAP1 nuclear-to-cytoplasm ratio for female HUVECs at each factor-level combination of shear stress and substrate stiffness. The colormap indicates regions of high (yellow), mid (green), and low (blue) values. Comparison surface plots combining the male and female plots together for median C) cell area, F) nuclear area, and I) YAP1 nuclear-to-cytoplasm ratio. Female plots are colored red and male plots are colored blue.

3.3. Main Effects

We next fit Equation (1) to model the response surfaces for male and female HUVECs to reveal the influence of each factor and their combinations. From the analysis of variance tables for the model fitting, all factors in Equation (1) were significant with p ≤ 0.0001 except for shear stress on nuclear area as visualized by Pareto charts of the standardized effects of each term in the model (Figure S2, Supporting Information). We also conducted three-way analysis of variance with post-hoc Šidák tests for multiple comparisons to indicate which corner points were statistically different from one another (Figure 6). For multi-factor models, to visualize the effect of each factor and their combinations on the response, it is useful to prepare main effect and interaction effect plots. Main effect plots (Figure 7) indicate the trends of individual factors while interaction effect plots (Figure 8) indicate the trends of pairwise combinations of factors. In general, the greater the slope, the greater is the influence of the factor on the response, the difference in slope for pairwise combinations, and the stronger is the interaction between the two factors.

Figure 6.

A,B,G) Pairwise comparisons of cell area, B,E,H) nuclear area, and C,F,I) YAP1 nuclear-to-cytoplasm ratio between the high and low levels for each factor. Each response was transformed by taking its natural logarithm to satisfy normality conditions for evaluation by three-way ANOVA and post-hoc Šidák multiple comparison tests. Comparisons were made without the center points. ****p < 0.0001, ***p < 0.001, **p < 0.01, *p < 0.05, ns p ≥ 0.05. Data presented as mean ± standard deviation of the natural logarithmically transformed response.

Figure 7.

Main effect plots. Plots of the median show the individual influence of A) sex, B) shear stress, and C) stiffness on cell area. Plots of the median show the individual influence of D) sex, E) shear stress, and F) stiffness on nuclear area. Plots of the median show the individual influence of G) sex, H) shear stress, and I) stiffness on YAP1 nuclear-to-cytoplasm ratio. Each factor had a statistically significant effect (p < 0.0001) except shear stress on nuclear area as determined from the ANOVA table for the model fit when transformed according to Table S4 in the Supporting Information. Center points (in red) can indicate nonlinearity in the response to the different factors if they are far away from the trend line between corner points and can indicate linearity if they are near the trend line between corner points.

Figure 8.

Interaction effect plots. Plots of the median show the interactive influence of A) sex and shear stress), B) sex and stiffness, and C) shear stress and stiffness on cell area. Plots of the median show the interactive influence of D) sex and shear stress, E) sex and stiffness, and F) shear stress and stiffness on nuclear area. Plots of the median show the interactive influence of G) sex and shear stress, H) sex and stiffness, and I) shear stress and stiffness on YAP1 nuclear-to-cytoplasm ratio. Each factor combination had a statistically significant effect (p < 0.0001) as calculated from the ANOVA table for the model fit when transformed according to Table S4 in the Supporting Information. Center points (in red) can indicate nonlinearity in the response to the different factors if they are far away from the trend line between corner points and can indicate linearity if they are near the trend line between corner points.

3.3.1. Sex

Independent of shear stress and substrate stiffness, female HUVECs spread more than male HUVECs (Figure 7A), female HUVEC nuclei were larger than those of male HUVECs (Figure 7D), and YAP1 was more present in the nucleus for female HUVECs than male HUVECs (Figure 7G). Of the main effects, sex had the greatest influence.

3.3.2. Shear Stress

Independent of sex and substrate stiffness, increasing shear stress increased cell spreading (Figure 7B). Shear stress displayed concave up nonlinear behavior yielding the smallest nuclei at the center point and similarly sized nuclei at the corner points. Shear stress did not have a statistically significant effect on nuclear area (Figure 7E). This is a result of the interaction effect between sex and shear stress visible from the surface plots (Figure 5D–F). This result reveals the convolution that can occur when male and female responses trend opposingly leading to a factor effect being considered not significant when its interaction (sex and shear stress) is significant. Lastly, YAP1 translocated to the nucleus more with increasing shear stress in a nonlinear manner. It was maximized by the center point (Figure 7H). This effect is largely driven by the influence of the male response, which changed greatly with increasing shear stress compared to the female response, which remained largely invariant.

3.3.3. Substrate Stiffness

Independent of sex and shear stress, increasing substrate stiffness increased cell spreading (Figure 7C), increased nuclear size (Figure 7F), and YAP1 translocation to the nucleus was weakly dependent on substrate stiffness, was nonlinear in behavior, and was maximized at the center point (Figure 7I). Like with shear stress, the influence on YAP1 translocation was largely driven by the male response.

3.4. Interaction Effects

3.4.1. Sex and Shear Stress

For the combination of sex and shear stress, the trend lines at each level were fairly parallel indicating little interactive effect (Figure 8A). However, the center points suggest a concave down nonlinear behavior in cell spreading for male HUVECs because the center point response was greater than either corner point response. At each corner point female nuclei were larger than male nuclei (Figure 8D). There was a slight inversion in the response between the sexes. Male nuclei were largely unaffected by shear stress while female nuclei were smallest at the center point suggesting concave up nonlinear behavior. At low shear stress females had more nuclear YAP1 than males while at higher shear stresses the difference was reduced (Figure 8G). Females had an invariant amount of nuclear YAP1 with shear stress whereas males had an increasing amount with increasing shear stress. Males showed a concave down response to shear stress indicated by the value at the center point being larger than at the corner points. At the center point, the ratio was invariant between the sexes.

3.4.2. Sex and Substrate Stiffness

The interactive effects were different for the combination of sex and substrate stiffness (Figure 8B). In this case, the trends were nonparallel. Sex exerted little effect for HUVECs cultured on stiffer substrates whereas sex demonstrated a strong effect for HUVECs cultured on softer substrates. The center point response was near that of high stiffness substrate suggesting that the loss of sex dependence diminished between 10 and 55 kPa. More interestingly, there existed an inverted response with substrate stiffness between the two sexes. Male HUVECs were smaller on softer substrates and larger on stiffer substrates while female HUVECs were the opposite, smaller on stiffer substrates and larger on softer substrates. Even one of the most routine responses, cell spreading, appears to be sexually dimorphic with substrate stiffness and illustrates the need to include sex as a biological variable when evaluating the influence of material properties. On stiffer substrates sex had little effect on nuclear size whereas on softer substrates female nuclei were larger than those of males (Figure 8E). The center points yielded the smallest nuclei for females suggesting nonlinearity. Females showed a strong concave up nonlinear behavior in response to substrate stiffness indicated by the dip at the center point whereas males showed a more linear increase in nuclear size with increasing substrate stiffness. The trends in YAP1 localization were identical to those for the combination of sex and shear stress (Figure 8H). Notably, males displayed nonlinear response in their cell area whereas females displayed nonlinear response in their nuclear area. This distinction is curious and requires further investigation as it was also present for the interaction between sex and shear stress.

3.4.3. Shear Stress and Substrate Stiffness

The sex independent combination of shear stress and substrate stiffness interacted to influence cell area indicated by nonparallel trend lines (Figure 8C). Shear stress had little effect on cells cultured on softer substrates (10 kPa) while it had a greater influence for cells cultured on stiffer substrates (100 kPa). The center point indicated nonlinearity in the interaction favoring stiffer substrates for increased spreading. Nuclear size was fairly invariant to substrate stiffness at low shear stresses (Figure 8F). While on stiff substrates, nuclear size increased at higher shear stresses and decreased slightly on softer substrates at higher shear stresses. The center point produced the smallest nuclei again suggesting a nonlinear response overall. On soft substrates, there was little effect of shear stress on the YAP1 nuclear-to-cytoplasm ratio while on stiff substrates the ratio increased with increasing shear stress (Figure 8I). As was true for many of the effects, the center points indicated nonlinear behavior.

4. Discussion

Using a full factorial design of experiment enabled the efficient exploration of both main and interactive effects between the three factors, sex, shear stress, and substrate stiffness. The MechanoBioTester platform was integral to this effort because it enabled the freedom to test combinations of shear stress and substrate stiffness without difficulty at the desired range of factor levels.^[39] Additionally, a full factorial design can be modified after the initial measurements. The factor-level combinations used in this set of experiments can be expanded to include star points and quadratic terms in the model to better resolve the nonlinear behavior of the responses.^[51] Moreover, the MechanoBioTester is a single system that also can test flow regime, cyclic stretch, hydrostatic pressure, and topography, not only shear stress and substrate stiffness. Having this functionality in a single system enables comparison of different mechanical stimulation combinations without introducing additional variability by changing stimulation systems. In a similar manner, the full factorial design model can be expanded to include other factors. Thus, the results from this observational set of experiments can be used to compare with future experiments using the MechanoBioTester system.

Endothelial cells exist in a very mechanically active environment and coincidentally their functionality is highly responsive to mechanical stimulation.^{[1,14,22,23,37,59,60]} Endothelial cells reorganize their cytoskeleton and cell membrane orientation and spreading in response to both shear stress and substrate stiffness.^[1,37] Our main effect findings agree with previous single factor mechanical stimulation studies that cell area increases for both increasing shear stress and increasing substrate stiffness.^[1,37] In culture it has been shown that laminar shear stress (>1.0 Pa) induces the production of nitric oxide and prostacyclin, two vasodilatory molecules.^[37,61–63] In fact, it was shown that nitric oxide production increased nonlinearly with increasing levels of shear stress from 0.2 to 1.2 Pa.^[63] Laminar shear stress enhances wound healing in the direction of flow and challenges it in the direction perpendicular to flow as a function of shear stress.^[64,65] Proliferation is reduced upon stimulation by shear stress and has been shown to reduce with increasing shear stress and substrate material.^[37,66] It is important to note that these studies were largely conducted on supraphysiologically stiff glass substrates. Similar effects have been observed for the influence of substrate stiffness on endothelial cell function. Migration has been shown to be enhanced on stiffer substrates in the range of 4 to 50 kPa.^[67] Proliferation increased with increasing substrate stiffness in the range of ≈2 to 21 kPa.^[68] Endothelial cell stiffness increased on stiffer substrates for substrates in the range of 0.1 to 10 kPa.^[69] Shear stress also exerted a stiffening effect on endothelial cells.^[70,71] Both stiffening effects are attributed to reorganization of the cytoskeleton as a consequence of changes in cell shape induced by the stimuli.^[69,70] Despite both shear stress and substrate stiffness having significant effects on endothelial cell function their combination has been scarcely investigated; despite, observed interactive effects.^[1] Using a cone-and-plate viscometer, bovine aortic endothelial cells were stimulated by both shear stress at 1.2 Pa and substrate stiffness at 2.5 and 10 kPa along with glass as a control substrate to investigate their effect on nitric oxide production and signaling.^[23] From this work, it was shown that nitric oxide production was enhanced for cells cultured on the compliant 2.5 kPa substrates compared to glass upon application of shear stress. In evaluating the temporal signaling for this response, comparison of the glass substrate and the 10 kPa substrate revealed no difference in ERK1/2 and nitric oxide synthase phosphorylation. Both phosphorylation events occurred later than for the 2.5 kPa substrate suggesting that this signaling might be more sensitive to more compliant substrates in the presence of laminar shear stress. Interestingly, no difference in cell area was found. We observed a significant difference in cell area between 10 and 100 kPa when sheared at 1.5 Pa for both males and females (Figure 6G). This discrepancy could be due to differences in the magnitudes for the two stimuli, differences in the cell type, or differences in the stimulation device, to name a few possibilities. In a separate study using a modified parallel-plate flow chamber, bovine aortic endothelial cells were stimulated by all combinations of 0.6, 1.2, 1.8, and 2.2 Pa shear stress and substrates with modulus of 0.1, 2.5, and 10 kPa to investigate their effect on the response by endothelial cells.^[22] This work demonstrated interactions between shear stress and substrates stiffness with cell area. It was suggested that stiffer substrates could mask the effect of shear stress in other words reducing sensitivity to shear stress. Our findings contend with this as we observed a return of sensitivity to shear stress with an increase in substrate stiffness (Figure 6D). This discrepancy could be due to differences in the magnitude of the substrate stiffness evaluated among other aforementioned differences with using different cell types and stimulation methods. Our low stiffness value of 10 kPa was their high stiffness value. It could be that additional changes occur that reverse the shear stress insensitivity observed at 10 kPa and is worth investigating in more detail. Overall, it is difficult to make comparisons and draw conclusions from previous work as the stimulation ranges largely differed. Our selection of the ranges for shear stress and substrate stiffness were intended to encompass a broad space to cue the field for locations to further investigate in more detail.

Absent entirely from these previous in vitro cell studies was biological sex. We have shown for the first time in vitro that the HUVEC response to combinations of mechanical stimulation is sex-specific—male and female HUVECs did not behave the same way. Both spreading and YAP1 localization were greater for females than males (Figure 7A,D,G). In static culture, endothelial cells have been shown to display sexual dimorphism for a range of functions including proliferation, vasoregulation, migration, and wound healing, tube formation and vascular endothelial growth factor stimulation, response to stressors of hyperoxia and serum starvation, secreted factors, and stimulation by sex hormones.^{[41,42,44,45,46,48–50,72–77]} Our main effect results for cell spreading agree with reports on the characteristics of male and female rat aortic and microvascular endothelial cells that found female endothelial cells spread more than male endothelial cells.^[46] It is important to note that this has immediate implications in terms of barrier function and wound healing ability, i.e., being able to cover a given area. It implies that less female endothelial cells would be necessary to cover a wound of a given area. Of note, functionally female HUVECs have been observed to proliferate and migrate faster, produce more nitric oxide synthase, the enzyme responsible for production of vasodilatory nitric oxide, and respond more efficiently with production of prostacyclin and prostaglandin E2 upon challenge by thrombin, a vasoconstrictor.^[43,49,76] These differences have significant implications for in vitro experiments used to evaluate biomaterials. We intentionally included a brief review of the effects of shear stress and substrate stiffness on endothelial cell function beyond those measured by us presently to illustrate that these features are all influenced by sex, as well. Because of these overlapping effects of biological sex and mechanical microenvironment it becomes clear that sex as a biological variable must be considered for in vitro biomaterials studies.

Only one other in vitro study has investigated the interaction of sex and shear stress on endothelial cell response. Lorenz et al. investigated the influence of 0.6 and 1.5 Pa shear stress on male and female HUVECs cultured on glass substrates.^[41] For those cells exposed to 0.6 Pa of shear, they had their entire transcriptome sequenced. It was reported that there were 1000+ more genes regulated by female HUVECs compared to males. This condition best approximates our low shear stress (0.5 Pa) and high substrate stiffness (100 kPa) stimuli combination for which we observed a notable sex difference in our measured responses (Figure 6A–C). From their transcriptomics analysis, expression differences for select markers (VCAM-1, eNOS, NOX-4, HO-1) were confirmed by RT-qPCR. When shear stress simulation was increased to 1.5 Pa, the differences in expression levels for the select markers between the males and females disappeared. This condition best matches our high shear stress (1.5 Pa) and high substrate stiffness (100 kPa) stimuli combination. Our results corroborate with this transition as we observed a loss of differences in our measured responses in moving from a low to high shear stress condition on our high stiffness substrate (Figure 6A–C). Our functional results from a broader range of stimulation conditions echo these findings for two levels of shear stress—endothelial cells display sex-specific phenotypes in response to mechanical stimulation. Most notably, within our range of stimuli conditions, male HUVECs were more sensitive to the mechanical stimuli than female HUVECs.

Two increasingly relevant molecular regulators of mechanobiological action are transcriptional co-factors Yes-associated protein 1 and transcriptional coactivator with PDZ-binding motif (TAZ). When active, YAP/TAZ translocate from the cytoplasm to the nucleus where the proteins complex with TEA domain family member transcription factors. Thereby the complex transcribes genes related to proliferation, apoptosis, and cell fate.^[78] YAP/TAZ translocation can be mechanically driven by both shear stress and substrate stiffness and it has been correlated to atherogenic and atheroprotective endothelial cell phenotypes.^[56–58] Previously shown for endothelial cells, YAP/TAZ translocation to the nucleus results from stimulation by oscillatory flow whereas it remains comparatively inactive in the cytosol from stimulation by unidirectional, laminar flow. Downstream of this signaling, more active YAP/TAZ stimulated by oscillatory flow promoted atherogenic behavior while less active YAP/TAZ stimulated by laminar flow promoted atheroprotective behavior of endothelial cells.^[57,79–81] For context, values for the YAP1 nuclear-to-cytoplasm ratio for laminar flow and for oscillatory flow were ≈1 and ≈4, respectively.^[57] Conversely, increasing shear stress from zero (no shear) to physiological levels of unidirectional, laminar shear stress promoted YAP1 nuclear translocation associated with a YAP1 nuclear-to-cytoplasm ratio changing from ≈1 to ≈1.5, respectively.^[82] This level of translocation was found to be necessary for lumen maintenance.^[82] Thus, it is likely that the relative amount of YAP1 activity dictates whether endothelial cells are either atherogenic or atheroprotective. Additionally, endothelial cells are adherent cells and thus always experience the properties of their substrate. However, these in vitro studies of YAP1 translocation in endothelial cells upon shear stress stimulation were conducted on supraphysiologically stiff glass substrates. In contrast, the endothelial cell basement membrane is significantly softer with moduli on the order of 10s of kilopascals not gigapascals.^[1,57,82] With this in mind, in general, it has been shown that stiffer substrates promote YAP1 nuclear translocation and more spreading whereas softer substrates promote its retention in the cytoplasm and less spreading.^[56,83] We observed YAP1 nuclear-to-cytoplasm ratios ranging from ≈1.2 to ≈1.6, which are in the range of those reported for laminar shear stress. We did observe an increase in ratio with increasing shear stress that agrees with those previous reports of YAP1 nuclear content increasing with the onset of shear stress compared to a no shear stress condition. Importantly, we saw that YAP1 activity in response to shear was mediated by the substrate in the case of male HUVECs. Conversely, little change in activity was observed for female HUVECs, which remained around the level previously reported for lumen maintenance.

The morphological differences we observed between male and female endothelial cells in response to the combinations of shear stress and substrate stiffness suggested there could be a commensurate difference in YAP/TAZ activity. Changes in cell spreading, nuclear spreading, and the arrangement of the F-actin cytoskeleton are key signals influencing YAP/TAZ activity.^[84–87] By plotting YAP1 nuclear-to-cytoplasm ratio versus cell spreading, two distinct groups appeared (Figure 9). One group had relatively higher activity and more cell spreading and one had lower YAP1 activity and less cell spreading. Male HUVEC YAP1 nuclear localization approached that of female HUVECs only when the combination of mechanical stimulation was above a threshold level for both shear stress (≥1.0 Pa) and substrate stiffness (≥55 kPa). These clusters illustrate that different complex mechanical microenvironments can lead to sex-specific phenotypes and sex invariant phenotypes.

Figure 9.

Plot of the median YAP1 nuclear-to-cytoplasm ratio versus the median cell area for each sex. Bubbles enclose the maximal interquartile area of the two measurables. The two male conditions clustering with the female conditions were those stimulated by the center point level of shear stress and substrate stiffness (1.0 Pa and 55 kPa) and the corner point with high levels for both shear stress and substrate stiffness (1.5 Pa and 100 kPa).

YAP1 nuclear-to-cytoplasmic localization changed with each factor-level combination of stimuli for male HUVECs whereas for female HUVECs it remained invariant. This sensitivity could translate into differences in health and disease. Female invariance to different levels of mechanical stimuli in terms of YAP1 activity could be a mechanism for the apparent age-related athero-protectiveness expressed by females compared to males. The atypical symptoms expressed by women with peripheral arterial disease could result from a loss of this invariance to mechanical stimuli at later stages of life that leads to a decline in endothelial function.^[38,88] In any event, this line of inquiry requires greater investigation. Additionally, from the sex independent interaction effect plots it became clear that the combination of shear stress and substrate stiffness on spreading and YAP1 localization (Figure 8G–I) was nonlinear. This implies that there exist optimal combinations that can maximize or minimize the cell response. This finding reiterates that mechanical stimuli cannot be investigated or applied in isolation. At the very least, when studying adherent cell types, the influence of substrate stiffness must always be included.

In the grand scheme of things, these results can inform the design of precision biomaterials.^[36,89,90] Because male and female HUVECs respond differently to shear stress and substrate stiffness and with a nonlinear dependency, there exists the combination of material properties and patient properties that can elicit the most beneficial cellular phenotype in vivo. For instance, the combination of 3D printed biomaterials and patient data could be used to achieve this end. Imagine a patient in need of a vascular graft or stent that could be 3D printed in such a way to spatially pattern the stiffness of the device to compensate for local flow patterns of the patient’s vasculature measured by 4D MRI so that the patient’s endothelial cells that reendothelialize the device are pushed toward a healthy, atheroprotective phenotype. Strategies like these are within reach but require investment to increase our understanding of sexual dimorphism in complex microenvironments.

5. Study Limitations

Though this exploratory study made insightful observations on sex-specific endothelial cell response to combinations of mechanical stimulation, it was not without limitations owing to its nature of being an in vitro assessment of cell function. We performed an observational study with statistics based on single cell observations from pooled male and female human umbilical vein endothelial cells from three male and three female donors conducted as three independent experiments at each factor-level combination. This gave us three pseudo-biological replicates because our measurements were made at the single cell level such that any given cell that was measured could be from any one of the three donor populations. Thus, any individual donor population could contribute equally to the total number of measurements made without any experimenter bias for a given donor population as we were blinded to the origin of any given cell after pooling. It is not uncommon for studies of sex difference in endothelial cells and other cell types to use single cell populations or single pooled populations for males and females when reporting sex differences.^{[46,49,75,91,92]} Pseudo-biological replicates in the form of pooling with single cell measurements were used as a compromise to experimental resources of time and cost to reduce a potential 90 experiments (3 replicates of 5 factor-level combinations for 3 male and 3 female populations) to 30 experiments (3 replicates of 5 factor-level combinations for 1 male and 1 female pooled population) while keeping sufficient technical replication.

As we begin to probe the influence of multiple factors in vitro a number of contributors should be considered and recognized. These may include the influence of molecular contributors such as past environments and epigenetic changes, donor demographics and medical history, donor genetic variability, the species and tissue origin of the donor cells, and the lack and inclusion of cell–cell interaction by self and other cell types.^{[11,43,93–100,101–103]} They also take the form of environmental contributors such as the choice of stimulating device, the choice of substrate material and extracellular matrix protein used for attachment, the collection of mechanical and chemical factors omitted from investigation, the range of stimuli included for investigation, the supplements included in cell culture media formulations, and the influence of sex hormones.^{[1,45,47,104–108]} Addressing all of these molecular and environmental contributors is untenable for the in vitro medium, thus the best course of action is to recognize these contributors and report in sufficient detail the information that is known and unknown. We concede that we are unable to provide detailed information on the donor because the cells we used were commercially purchased and information is unavailable from the vendor beyond cell type, donor age, and donor ancestry. We have included this information for each cell population in Table S5 in the Supporting Information. We have included information on the past environments the cells experienced in the Experimental Section, which detail the culture substrate and media formulations used. Relevant to studies considering sex is the consideration of sex hormones. We used a commercially prepared, serum-free, defined media for our stimulation experiments to mitigate the influence of different supplemented serum and thus hormone compositions across bottles of media.^[104,106] Additionally, the media was free of phenol red because phenol red can have estrogenic effects.^[109,110] We recognize that in the context of endothelial cells there are sizable cell–cell interactive effects in vivo among others those related to interactions with vascular smooth muscle cells and macrophages and thus our results will likely differ when those other cell types are included.^[99–101] Similarly, endothelial cells display tremendous heterogeneity in their behavior depending on their tissue of origin.^[95–98] Such effects have been observed to interact with biological sex, as well.^[46] Consequently, it is necessary to expand upon our findings by evaluating not only HUVECs, but other endothelial cell types derived from other vessels such as the aorta, carotid artery, or microvasculature. In our study, we used the MechanoBioTester platform for its versatility to produce combinations of mechanical stimuli that said other devices exist for stimulating by only shear stress and substrate stiffness.^{[22,23,37,39,40]} When stimulated, our cells were adherent to a type I collagen-coated polydimethylsiloxane substrate. Extracellular matrix protein and underlying material can influence cell response, and these could both contribute to the responses we observed by influencing the mechanotransduction machinery of the cell.^[1] Overall, each of these limitations are additional avenues to pursue in terms of their combinatorial effect on endothelial cells and cells in general and warrant their further investigation.

6. Conclusion

From this exploratory study, it was shown that in complex mechanical microenvironments, HUVECs displayed sex-specific phenotypes. This fact has far reaching implications for treating and preventing cardiovascular disease, understanding mechanobiology, and developing precision biomaterials. Yet, much still remains to be understood and new questions need answers. Future work should look to understand in more detail YAP1 sensitivity and lack thereof to mechanical stimulation by male and female endothelial cells from different tissues.

7. Experimental Section

Cell Culture:

Male and female HUVECs were cultured and pooled. Three populations each of male and female HUVECs were purchased from Cell Applications (Table S5, Supporting Information). Each population was expanded using Endothelial Growth Medium sold without phenol red (Cell Applications) to passage 3 and cryopreserved. The phenol red-free version of the media formulation was used because phenol red is known to have weak estrogenic effects.^[109–111] Then male-only cultures of pooled HUVECs and female-only cultures of pooled HUVECs were prepared by combining an equal number of cells (counted by hemocytometry) from each individual population at passage 4 and cryopreserved (Figure S4, Supporting Information). All cultures were grown on tissue culture polystyrene in a humidified cell culture incubator kept at 37 °C and 5% CO₂. Media was exchanged every 2–3 days.

Sex Validation:

The sex of each population of male and female HUVECs was confirmed by PCR. During expansion of each population, ≈500 000 cells were pelleted, lysed in DNA/RNA Lysis Buffer (Zymo Research), and frozen at −20 °C. From the lysate, genomic DNA (gDNA) was purified using a Quick-DNA/RNA Microprep Plus Kit (Zymo Research) using molecular-biology grade reagents and consumables as required. Purified gDNA was stored at −80 °C. gDNA concentration and purity was measured spectrophotometrically using a NanoDrop ND1000 (Table S6, Supporting Information). Sex-specific sequences of gDNA were amplified by PCR according to a modified protocol by Finch, Hope, and van Daal.^[54] The PCR reactants followed the manufacturer’s protocol for One-Taq Quick-Load 2X Master Mix with Standard Buffer (New England Bio-labs), 100 ng gDNA, 2 μL primers (1 μL of each primer at 10 × 10⁻⁶ m), 25 μL of OneTaq, and the remaining volume was ultrapure, molecular-biology grade water to bring the reaction volume to 50 μL. The protocol uses two primer sets (Table 2) to amplify in a single reaction a region of the SRY gene, specifically the region for TDF (males only), and a region of the HLA-DQα gene, an internal control found on chromosome 6 (both males and females). Primers were purchased from Integrated DNA Technologies and reconstituted and diluted with ultrapure, molecular-biology grade water. The PCR mix was subjected to a 30 s initial denaturation at 94 °C followed by 35 cycles of 94 °C for 30 s, then 60 °C for 60 s, and then 72 °C for 60 s in a CFX Connect Thermocycler (BioRad). The PCR products are short DNA sequences of 139 bp (SRY) and 239/242 bp (HLA-DQα). DNA gel electrophoresis was used to evaluate the presence of the two reaction products. A 2% agarose gel was prepared by mixing sufficient certified low-range ultra-agarose (BioRad) with 1× TAE buffer (Fisher Scientific) and heating in a microwave to dissolve the agarose powder. Upon adequate cooling, ethidium bromide (Sigma Aldrich) was added to the agarose solution to give a final concentration of 500 ng mL⁻¹. Then the solution was poured into a Wide Mini-Sub Cell GT gel box (BioRad) to cast the 2% agarose gel. After gelation, the gel box was filled with 1× TAE buffer to sufficiently cover the gel. Then 5 μL of each PCR product was mixed with 1 μL of 6× DNA Gel Loading Dye (ThermoScientific) and was pipetted into its own lane in the gel. 7.5 μL of Quick-Load (R) 100 bp DNA Ladder (New England Biolabs) was also pipetted into a lane in the gel. The gel was run at 70 V (10 V cm⁻¹) for 30 min and then imaged on a PhotoDoc-It transilluminator (UVP).

Culture System:

Mechanical stimulation experiments were conducted using the MechanoBioTester system described previously.^[39] The system uses a custom-designed cell culture chamber that can be used to stimulate cells by decoupled combinations of fluid flow, cyclic stretch, hydrostatic pressure, and substrate properties. Male and female HUVECs were mechanically stimulated by combinations of fluid flow and substrate stiffness for 24 h. In all experiments, chambers were fabricated to have a cell culture region (CCR) with the desired stiffness level following the protocol described in CCR substrate: polydimethylsiloxane elastomer (PDMS).^[39] Chamber CCRs were treated for cell adhesion following the protocol described in CCR substrate treatment for cell adhesion using rat-tail type I collagen diluted to 0.3 mg mL⁻¹ with sterile deionized water (Corning).^[39] In all experiments, male or female pooled HUVECs were seeded to confluency (≈30 000 cells cm⁻²) at passage 4 and allowed to attach for ≈6 h prior to chamber sealing with ReproRubber Thin Pour (ReproRubber). After complete chamber sealing, the chamber was flushed with PBS and then refilled with fresh endothelial growth medium without phenol red (cell applications). Before stimulation experiments, the cells were cultured overnight in the chamber kept in a humidified cell culture incubator at 37 °C and 5% CO₂. All stimulation experiments were conducted using endothelial serum-free defined medium (cell applications). The media viscosity was increased with polyvinylpyrrolidone (PVP) (P0930, Sigma Aldrich) to make a 10% (w/v) PVP-media solution. The viscosity of the PVP-media solution was measured at 37 °C using a Brookfield viscometer and ranged from 4.25—4.60 cP, which is in the range of human blood (Figure S5, Supporting Information). Accounting for the measured PVP-media viscosity and chamber dimensions, the flow rate was adjusted to achieve the desired shear stress for any given flow experiment according to Equation (2)

$$\tau =\frac{6Q\mu }{W{H}^2}$$ (2)

where τ is the wall shear stress, Q is the volumetric flow rate, μ is the fluid dynamic viscosity, W is the channel width, and H is the channel height. The seeded chambers were connected to the MechanoBioTester system and the flow rate was incrementally increased by 16.5 mL min⁻¹ every 30 min up to the flow rate that gave the desired shear stress. Once the target flow rate was reached, stimulation was sustained for 24 h. Following stimulation, chambers were disconnected, rinsed with PBS three times, and the cells within were subsequently analyzed.

Immunocytochemistry:

To evaluate YAP1 localization and F-actin cytoskeletal organization, after experimentation cells were immunostained. The cells were fixed with 10% formalin (Sigma Aldrich) for 15 min and rinsed with PBS three times. Cells were then permeabilized for 10 min with 0.1% Triton X-100 (Fisher Scientific) in PBS and rinsed with PBS three times. Cells were blocked for 30 min with 1% bovine serum albumin (Fisher Scientific) in PBST (0.1% (v/v) Tween 20 (Fisher Scientific) in PBS). Following blocking, cells were incubated with anti-YAP1 [EP1674Y] monoclonal antibody (ab52771, abcam) diluted 1:500 in l% bovine serum albumin in PBST overnight at 4 °C. Cells were then rinsed four times with PBS. Afterward, cell cytoskeletal F-actin was stained with Texas Red labeled phalloidin (Invitrogen) diluted in 1% bovine serum albumin (Fisher Scientific) in 0.1% Tween 20 (Fisher Scientific) in PBS following the manufacturer’s instructions. In this same solution, anti-rabbit AlexaFluor 488 tagged secondary antibody (ab150077, abcam) was added at a dilution of 1:500. Incubation with the phalloidin and secondary antibody was for 1 h at room-temperature followed by five rinses with PBS. The CCR region of the chamber was then excised from the chamber and mounted using ProLong Diamond Antifade Mountant with DAPI (Invitrogen) onto a glass slide (ThermoFisher Scientific).

Morphometrics and YAP1 Localization:

Immunostained cells were imaged using a Zeiss Axios Observer inverted microscope with both wide-field (Colibri 7 excitation source) and confocal (LSM 900) capabilities. The phalloidin stained cellular F-actin and the DAPI stained cell nucleus were imaged in widefield. 9 to 15 images at random positions were taken for each experimental replicate. A simple, quantitative metric for measuring YAP/TAZ activity is to evaluate the ratio of YAP1 in the nucleus to YAP1 in the cytoplasm, which can be determined by immunocytochemical staining.^[112] Immunostained cells were imaged by confocal microscopy. 9 to 15 images at random positions were taken for each experimental replicate. For all imaging, locations were specific to the 1 cm by 2 cm region of the MechanoBioTester cell culture region. Images were taken randomly within a concentric 0.5 cm by 1 cm rectangular region to avoid including those cells at interface between of the cell culture region and the rest of the flow channel. In each image, every cell was manually outlined as a region of interest. Then using the colocalization tool in Zeiss Zen Blue microscopy software (version 3.1), the average cytoplasmic YAP1 intensity and nuclear YAP1 intensity was determined for each region of interest (each cell). The nuclear region was defined from the DAPI channel whereas the cytoplasmic region was defined from the YAP1 channel. Nuclear intensity was defined as the colocalization of the DAPI channel and YAP1 channel measured with respect to the YAP1 channel within the region of interest (manually outlined cell). The cytoplasmic intensity was defined as the non-colocalized pixels of the YAP1 channel within the region of interest (manually outlined cell). The ratio of YAP1 in the nucleus to the cytoplasm for each cell was calculated as the ratio of the average nuclear YAP1 intensity and the average cytoplasmic YAP1 intensity. Cell area and nuclear area were calculated as the total YAP1 channel within the region of interest and the colocalization of the DAPI channel and YAP1 channel within the region of interest, respectively.

Statistical Analysis:

Statistical analyses were conducted in Minitab 19.2020.1. Using the Minitab design of experiment feature, a 2^k full factorial design with one categorical (sex) and two continuous (shear stress and substrate stiffness) factors with center points for the continuous factors could be designed and analyzed. The responses were automatically Box-Cox transformed to satisfy normality conditions for fitting to a saturated linear model with center points (Equation (1)). An alpha value of 0.05 was used to determine statistical significance of each factor combination. Analyses for both morphometrics and YAP1 localization were conducted on a per cell basis for measurements pooled across three independent experimental replicates. Main effect and interaction effect plots were prepared by combining the measurements on a per cell basis for each factor-level combination into a single distribution as appropriate for each specific effect, e.g., to calculate the point on the main effect plot for cell spreading for males and for females, the measurements on a per cell basis that were for males regardless of shear stress or substrate stiffness level were combined and in a similar manner those for females were combined to yield two distributions. This process was then carried out in the same way for each other factor and pairwise factor combination. Pairwise comparisons between corner points were evaluated for statistical significance using a three-way analysis of variance with Šidák multiple comparison tests. For all statistical tests, a p value less than 0.05 was considered statistically significant. Violin plots, main effect plots, and interactive effect plots were prepared using GraphPad Prism 8.1. Linearly interpolated surface plots were prepared in MATLAB R2020b.

Data Availability Statement

The data that supports the findings of this study are available in the Supporting Information of this article.