Structural and mechanical remodeling of the cytoskeleton studied in 3D microtissues under acute dynamic stretch

When mechanically stretched, cells cultured on 2D substrates share a universal softening and fluidization response that arises from poorly understood remodeling of well-conserved cytoskeletal elements. It is known, however, that the structure and distribution of these cytoskeletal elements are profoundly influenced by the dimensionality of a cell’s environment (ie. on a 2D surface vs. within a 3D matrix). Therefore, in this study we aimed to determine whether cells cultured in a 3D extracellular matrix also follow the same softening response and to link this mechanical change to direct evidence of cytoskeletal remodeling. To achieve this, we developed a new high-throughput approach to measure the dynamic mechanical properties of cells and allow for sub-cellular imaging of physiologically relevant 3D microtissue cultures. We found that fibroblast, smooth muscle and skeletal muscle microtissues strain softened but did not fluidize, and upon loading cessation, they fully regained their initial mechanical properties. These responses required a filamentous actin cytoskeleton, and were mirrored by changes to actin remodeling rates, and direct visual evidence of actin depolymerization during stretching and repolymerization after stretch cessation. On the other hand, the response could not be attributed to either remodeling of microtubules or myosin motor activity. Our new approach for assessing cell mechanics has linked behaviors seen in 2D cultures to a soft 3D extracellular matrix, and connected visual remodeling of the cytoskeleton to changes in mechanical properties at the tissue-level. Significance Statement With every breath and movement, cells in our body are subjected to mechanical forces. These forces are key regulators of normal development and function, as well as disease progression. To understand how cells “feel” mechanical cues in their microenvironment, we have previously relied on two-dimensional experimental approaches and often assessed single cells in isolation. Here, we present a novel lab-on-chip device, which enables simultaneous mechanical stimulation and sub-cellular imaging of three-dimensional multi-cellular microtissues. In this article with this device, we quantitatively linked force-induced mechanical changes in microtissues to specific molecular remodelling pathways in the cytoskeleton. The approaches and insights presented in this study will deepen our understanding of the mechanobiological pathways governing tissue development and function in health and disease.


Introduction
With every breath, heartbeat and movement, cells in our body experience mechanical stretch, which in turn, creates continually unstable forces at focal adhesions, across the cell membrane, along cytoskeletal filaments and through the nucleus 1,2 . In a cell, these forces direct functional and phenotypic behaviors by generating conformational changes, and thereby, altering ligand-receptor affinities 1,2 . Importantly, this ability of cells to feel and adapt to mechanical forces has been linked to crucial events in normal development and function, as well as disease progression, including bone, muscle, heart and lung disorders, and cancer 3,4 .
In particular, the well-conserved structural elements that make up the cytoskeleton of eukaryotic cells are in themselves mechanosensitive; in response to dynamic stretch, the cytoskeleton softens (decreased elasticity) and becomes more fluidlike [5][6][7][8] . Then upon stretch cessation, it slowly regains its stiffness and resolidifies 7 .
Currently the molecular mechanism(s) behind strain softening remains unclear as there is limited visual-based evidence quantifying cytoskeletal remodeling following cyclic stretching [9][10][11] . Furthermore, softening and fluidization has been observed in response to deformation at the subcellular 5,6 and single cell levels 7,8 , however the extent of the response remains poorly understood at the tissue-level [12][13][14][15] . Nevertheless, in the body, this response has been linked to the maintenance of airway caliber 12,13 and the regulation of blood pressure 16 , but for unknown reasons, it is absent in certain pathological disorders. For example, unlike in healthy lungs, stretch from a deep inspiration does not dilate asthmatic airways 17 .
Most of our knowledge of cell mechanics and how cells respond to mechanical forces has been generated by growing cells on rigid, flat surfaces. Yet it is known that the physical environment in which a cell is grown alters its mechanical properties and behavior. For example, cells grown on stiff substrates tend to have their actin cytoskeleton arranged into dense stress fibers, and are stiffer, more solid-like and under greater pre-stress when compared to cells on softer substrates [18][19][20][21] . In addition to matrix stiffness, it is suspected that the mechanical behavior of cells may be further altered by the dimensionality of their environment. In support of this growing hypothesis, culturing cells on a 2D substrate vs. within a more physiologically relevant 3D matrix fundamentally changes the distribution and structure of the cytoskeleton by forcing unnatural apical-basal polarity of adhesion complexes 22 . The difference between a rigid, flat, petri dish and a soft 3D extracellular matrix may also explain observed disparities in cellular behavior, and the loss of efficacy in costly clinical trials that often occurs when pharmaceutical treatments are developed using conventional 2D cell culture techniques [23][24][25][26] . Thus, there exists a need for new high-throughput cell culture techniques capable of probing cell mechanical behavior while maintaining a physiologically relevant soft 3D environment.
To address this need, techniques that allow assessment of the mechanical behavior of cells within reconstituted 3D collagen gels have been a keen interest to the fields of mechanobiology, pharmacology, and tissue engineering 27 . These methods have furthered our understanding of how tissue level forces are collectively generated by cells [28][29][30] , and their mechanical behaviors, including strain softening 31,32 . The centimeter scale of these bulk gels, however, limits throughput, causes imaging difficulties, and produces a high diffusive barrier for nutrients.
To overcome the limitations of bulk 3D cell cultures and to study the rapid dynamics and force generation during contractility, Legant et al. (2009) developed Microfabricated Tissue Gauges (microtissues) 33 . In the microtissue model, cells are cultured within a matrix composed of collagen and form around pairs of flexible vertical cantilevers into an array of dense, organized structures comparable to ex vivo tissue.
High-throughput tensile force measurements can then be calculated from the visible deflection of the cantilevers. More recently, investigators fixed a magnetic microsphere to one of the cantilevers in each microtissue well, and with magnetic tweezers stretched one microtissue at a time for quasi-static stiffness measurements 34,35 . The limitations in experimental throughput and actuation range of magnetically driven devices were addressed by our recently published Microtissue Vacuum-Actuated Stretcher (MVAS) 36 .
In that publication, the MVAS allowed for high-throughput visualization of cellular remodeling during stretching due to a mostly planar deformation and following chronic (several days) conditioning We now present a new microtissue stretcher, the MVAS-force, which enables measurements of tensile force and dynamic stiffness. In contrast to our previous design, only one of the cantilevers in the MVAS-force is actuated through a regulated vacuum pressure and forces are measured simultaneously from the passive bending in the other cantilever. In this article, this new approach allowed us for the first time to assess the mechanical properties of microtissues during dynamic loading and upon loading cessation, and to link the changes in mechanics to sub-cellular remodeling using responses to pharmacological treatments and by directly imaging the cytoskeleton. The findings that can be gained from our approach on a cell's ability, or impaired ability to sense mechanical forces are critical to understand pathways of development, normal function and disease progression in the body

Microtissue Morphology
Within the MVAS-Force, 3T3 fibroblast cells self-assembled around the cantilevers into dense, highly organized, three-dimensional constructs that morphologically resemble tissue. Top-down and cross sectional views of a fully compacted, representative microtissue are shown in fig. 1c. As been shown previously 33,36,37 , the cells compacted the collagen matrix away from the bottom and sides of the well into a tissue freely suspended around the tops of the cantilevers. The average microtissue thickness measured at its center after four days was 97 ± 2 µm (n=5) and was qualitatively uniform along the longitudinal axis.
Maximum intensity projections with orthogonal slices and centrally located magnified views of F-actin and cell nuclei within a representative microtissue at four days are shown in fig. 1d. Actin was highly polymerized into dense stress fibers that oriented with the longitudinal axis of the microtissue. The cell nuclei were also mostly aligned with the microtissue and evenly distributed in three dimensions.

Microtissues strain soften
It has been widely reported that acute dynamic stretching changes the mechanical properties of cells grown on 2D surfaces; they become softer (decreased elasticity) and more fluid-like (increased phase lag) 7,11,38 . In that regard, we started by investigating whether or not 3D microtissues composed of 3T3 fibroblasts share this behavior by assessing dynamic mechanical properties under progressively larger strains at 0.25Hz.
Unlike previously published findings on cells in 2D culture, where softening is accompanied with a more fluid-like behavior (or fluidization) 7 , the phase lag of microtissues decreased with strain amplitude (linear regression: R 2 = 0.95, p<0.01) ( fig.   2b), indicating a greater amount of energy stored for a given amount dissipated at higher strains. Therefore in contrast to cells in 2D, microtissues become more elastic-like as they soften.
We have shown that, as with cells in 2D culture, 3D microtissue cultures strain soften. Depolymerization of actin filaments 7,[9][10][11] and perturbing of myosin motor binding 12,13,[39][40][41] are previously hypothesized mechanisms for cells in 2D culture. On the other hand, the involvement of microtubules has been largely overlooked despite their contributions to overall cell mechanics [42][43][44][45] and their dynamic instability 46 . Accordingly, we investigated the roles of these three cytoskeletal proteins to microtissue strain softening. We begin with examining the involvement of actin microfilaments.

Softening is actin dependent
To assess the role of the actin cytoskeleton in strain softening, f-actin was depolymerized with Cytochalasin D (CytoD). As expected, CytoD treatment reduced the resting tension, stiffness and phase lag (SI 2). Importantly, CytoD treatment also muted the softening response (N=16 repeated measures t-test, p<0.001) ( fig. 3b). In fact the stiffness of CytoD treated microtissues did not change with strain amplitude (repeated measures t-test, P>0.05). Upon stretch cessation, CytoD treatment also prevented tension recovery; further demonstrating that CytoD treated microtissues do not soften ( fig. 3c).
These results indicate that strain softening is dependent upon changes to a densely polymerized actin cytoskeleton.

Stretch remodels and depolymerizes actin
We have shown that f-actin is required for strain softening in microtissues. To We have shown that oscillatory stretch increases remodeling of actin filaments in living cells in 3D cultures. To investigate whether remodeling arose simply from organizational changes or depolymerization/repolyermization of filaments, microtissues were fixed and stained immediately following various durations of stretching at 9% strain. Representative images, average heat maps and average f-actin expression per cell ( fig. 5a,b,c, respectively) all indicated that f-actin rapidly depolymerized with oscillatory stretching (N>14).
To show that f-actin also repolymerizes following stretch cessation, microtissues were fixed and stained after various durations of recovery following five minutes of stretching. Average heat maps and f-actin expression per cell ( fig. 5b,d) show complete recovery to initial expression values (t-test P>0.05) (N>11). Although our time resolution of f-actin expression was poor and the uncertainties are large, the rate of f-actin recovery appeared to be within the same order of magnitude (tens of seconds) as the rate of tension and stiffness recovery, suggesting that the mechanical measurements reflect actin repolymerization.

Myosin and microtubules do not contribute to strain softening
We have identified that actin filaments play a major role in the strain softening response of 3D microtissues, however, the mechanical behavior of cells [42][43][44][45]47 and microtissues (SI 2) is also highly dependent upon myosin activity and microtubules. To assess the contribution of myosin and microtubules to strain softening, we examined the response following myosin inhibition with blebbistatin (Bleb) and microtubule depolymerization with nocodazole (Noco).
Myosin inhibition with blebbistatin decreased microtissue stiffness (N=10, repeated measures t-tests, P<0.05) and prestress (p<0.01) (SI 2). Myosin inhibition, however, did not affect strain softening (N=10, repeated measures t-test, P>0.05). As expected, it did reduce the tension recovery ( fig. 6c) because of the decrease in prestress that unsurprisingly accompanied myosin inhibition. However, as there was no change in the rate of recovery, myosin was not likely responsible for the recovery following softening (SI 1). Although it is possible that there was incomplete inhibition of myosin motors, it is unlikely according to blebbistatin's measured dose-response curve (SI 3).
Moreover, even with incomplete inhibition, one would still expect a decrease in the softening response if myosin were responsible.
In keeping with the hypothesis that microtubules are mainly compressive elements that oppose acto-myosin activity 42

Discussion
We aimed at assessing the response to acute oscillatory stretch in cells grown in conditions that mechanically and biologically recapitulate the 3D environment that a cell would experience were it in the body. In order to fulfill this goal, we developed a novel approach to allow high throughput measurements of dynamic mechanical properties and direct visualization of the cytoskeleton in 3D microtissue cell cultures. Our approach consists of an array of vacuum-driven actuators to stretch microtissues and optically tracked force-sensors to measure their mechanical behavior. Advantages and limitations of our approach are summarized in SI 5. In using our approach, we showed that microtissues soften under dynamic stretching through actin depolymerization. These results are further discussed the sections that follow.

Microtissues soften under dynamic loading
We showed that the prestress and storage modulus of living 3D microtissue cultures composed with three different cell types decrease with dynamic strain amplitude.
This finding agrees well with previous reports in cells in 2D culture 7,38,48 and ex vivo tissue strips 12,13 . Strain softening has been identified as an important mechanism for maintaining homeostasis throughout the body. For example, it may explain how a large tidal stretch from a deep inspiration can open contracted airways in healthy lungs 12,13 and could contribute to the regulation of blood pressure in arteries 16 .
In addition to softening, cells in 2D have long been reported to exhibit a more fluid-like behavior when stretched [49][50][51] . More recently this fluidization response of cells has been associated with that of a class of inert materials called soft glasses (eg. foams, dense emulsions, pastes and slurries) 7 . In contrast to these reports and the soft glassy rheology hypothesis, skeletal muscle and fibroblasts microtissues actually became more elastic-like with greater strain amplitudes, and smooth muscle microtissues did not phase transition. Although this apparent contradiction could arise from environment differences between growing cells our 3D microtissues and past reports done in 2D culture, it is more likely that the fluidization response was hidden by the elasticity of extracellular matrix. In the literature strain softening exists in a paradox with a large number of studies reporting strain stiffening and actin reinforcement in response to stretch 57 36 . Importantly in those investigations quasi-static stiffness measurements and f-actin expression were evaluated following loading and after a period much greater than the time-scale of stiffness recovery and actin repolymerization we report here. That said, over our relatively short experimental time, we did not observe any differences between initial and fully recovered stiffness, prestress and f-actin expression measurements.

Stretch depolymerizes actin in microtissues
Although strain softening of cells and tissues has been widely reported 7,12,13,38,48 , the molecular mechanism(s) behind this response remains unclear. Here we investigated the contributions of actin microfilaments, myosin motors and microtubules.
Firstly, in a cell, the actin cytoskeleton is a filamentous network that gives the cell its shape and opposes tensile forces. In 2D culture, stretching of cells has been reported to depolymerize actin filaments [9][10][11] . Our results in 3D cultured microtissues agree with those observations. We found that 1) f-actin was necessary for strain softening and the recovery response; 2) actin remodeling in living cells increased with stretch; 3) shortterm stretch lowered f-actin expression; and 4) upon stretch cessation, f-actin expression recovered along the same timescale as tension and stiffness recovery. These findings strongly suggest that strain softening, at least in part, arises from actin depolymerization.
Secondly, myosin motors regulate the mechanical behavior of cells by generating tension through crosslinking and actively pulling on actin filaments. Furthermore, strain softening in reconstituted actin-myosin networks has been attributed to disruption of myosin crosslinks [39][40][41] . Perturbing of the binding of myosin has also been implicated in the softening response in airway tissue strips 12,13 . In contrast, we found softening in microtissues was invariant on myosin activity and that there was no change to the rate of the recovery response following strain cessation. This strongly suggests that myosin has no role in the softening response of cells in 3D culture.
Lastly, although our understanding of the role of microtubules in cell mechanics is still being refined 70 , it is thought that they act as compressive struts to oppose actinmyosin contractility, as in tensegrity architecture [42][43][44][45]

Conclusions
In this article, we presented a new high-throughput approach for both assessing dynamic cell mechanics and for visualization of remodeling at the sub-cellular level in response to stretch within physiologically relevant 3D microtissue cultures. Our approach offers the ability to link behaviors observed in 2D culture to cells within a soft 3D matrix comparable to human tissue, and to connect visual remodeling of the cytoskeleton to changes in mechanical properties. In that regard, we found that fibroblast, smooth muscle, and skeletal microtissue cultures all share a conserved softening response when dynamically stretched and recovery following stretch cessation. Furthermore, by directly quantifying cytoskeletal remodeling, softening of microtissues appeared to arise from rapid actin depolymerization. This suggests that actin microfilaments are sensors of mechanical stretch in cells, and in turn, form a feedback loop to control the mechanical behavior of tissues. The ability of cells to feel and react to mechanical stimuli from their environment is an important mechanism for maintaining homeostasis in the body and is a critical aspect to fully understand many pathological disorders.

Device design
Our original MVAS device 36

Microtissue fabrication
Microtissue fabrication was performed as described previously 33

Force measurements
Microtissue mechanics were deduced from the visible deflection of the forcesensing cantilever while under dynamic loading at 0.25Hz (movie 1). Prior to measurements, microtissues were preconditioned until subsequent loading loops were superimposable. All measurements were completed at 37 o C and 5% CO 2 . Microtissue strain, ε, was defined as the percent change in the length between the innermost edges of the tops of the cantilevers (equation 1).
The phase lag, δ, between force and strain was defined as the difference in the phase angles, Φ, at the oscillatory frequency (equation 2).
The storage, k', microtissue stiffness was defined as the ratio of the magnitudes of the Fourier Transforms of force and strain at the oscillatory frequency multiplied by the cosine of the phase lag between force and strain (equations 3).
k' describes the amount of energy that is elastically stored for a given deformation, and δ describes the ratio of energy dissipated to energy stored where in purely elastic samples tan(δ)=0 and in purely viscous samples tan(δ)=inf.
The tension offset or prestress, T o , was defined as the magnitude of the Fourier transform of the microtissue force at 0Hz minus the half of the peak-to-peak magnitude of the Fourier transform at 0.25Hz (equation 4).
The noise floor for calculating microtissue mechanics is characterized in SI 6.
To assess the response of microtissue mechanics to stretch, measurements were taken at progressively higher strains. After completing measurements at the largest strain, the recovery response was measured by promptly decreasing the strain amplitude.
Stiffness recovery was measured by performing Fourier transforms on intervals spanning three loading cycles.
To assess the role of individual cytoskeletal proteins in contributing to the mechanical properties of microtissues and the strain softening behavior, measurements were taken following 20 minute incubations with either 10µM nocodazole (Noco), a microtubule polymerization inhibitor, 5µM blebbistatin (Bleb), a myosin-II inhibitor, or 10 µM cytochalasin D (CytoD), an actin polymerization inhibitor. As mechanical properties can vary between microtissues, each microtissue was compared to its own pretreatment value where indicated. To prevent crossover in response from multiple drugs, only a single treatment was administrated to a microtissue.

Quantification of cytoskeletal remodeling and polymerization
Images were acquired on a TiE A1-R laser scanning confocal microscope (LSCM) (Nikon) with standard LSCM configurations using appropriate laser lines and filter blocks.
To assess actin and microtubule remodeling in living microtissues in response to stretch, cells were loaded with either 0.1µM SiR-actin or SiR-tubulin with 1µM verapamil 6-12 hours before imaging. Z-stacks were taken before and following 4 seconds (one stretch), one minute and five minutes of static resting and then ~9% stretching at 0.25hz. Imaging was completed with a 60x 1.2NA water immersion objective to give a centrally located field of view of 212x106µm (1024x512 pixels). Zstacks were flattened by integrating slices, divided into sub images with a size of 100x100 pixels with 10-pixel spacing, and compared with cross-correlation. The correlation coefficient is a measure of how closely images matched before and after a given condition, and thus is inversely proportional to the amount of remodeling (ie. a low correlation coefficient corresponds to a high degree of remodeling).
To assess f-actin expression and microtubule polymerization, microtissues were fixed in situ with 3.5% paraformaldehyde for 15 minutes and permeabilized with 0.5% Triton-X for 5 minutes. Microtissues were left in blocking buffer (5% FBS in PBS) for 40 minutes. Microtubules were labeled with α-tubulin primary antibody produced in mouse (Sigma, T6074) and a rabbit anti-mouse IgG secondary antibody conjugated to Alexa Fluor 488 (Invitrogen, A11059). The actin cytoskeleton was stained with Alexa Fluor 546 Phalloidin (Fisher, A22283), and the nuclei were stained with DAPI (Fisher, D1306). To quantify f-actin expression per cell, Z-stacks were flattened by integration, averaged and normalized to DAPI fluorescence. Microtubule polymerization was quantified with the same method except images were first thresholded to remove any signal from nonpolymerized tubulin.

Data analysis and statistics
All numerical data are presented as mean ± standard error. Statistical tests as described in the results were performed using Originlab 8.5 (Northampton, MA), with p<0.05 considered statistically significant.