Physical confinement alters sarcoma cell cycle progression and division

ABSTRACT

Tumor cells experience physical confinement on one or multiple axes, both in the primary tumor and at multiple stages during metastasis. Recent work has shown that confinement in a 3D spheroid alters nucleus geometry and delays cell division, and that vertical confinement impairs mitotic spindle rounding, resulting in abnormal division events. Meanwhile, the effects of bi-axial confinement on cell cycle progression has received little attention. Given the critical role of nuclear shape and mechanics in cell division, we hypothesized that bi-axial physical confinement of the cell body and nucleus would alter cell cycle progression. We used sarcoma cells stably expressing the fluorescence ubiquitination cell cycle indicator (FUCCI), along with fibronectin-coated microchannel devices, and explored the impact of bi-axial physical confinement on cell cycle progression. Our results demonstrate that bi-axial physical confinement reduces the frequency of cell division, which we found to be attributed to an arrest in the S/G2/M phase of the cell cycle, and increases the frequency of abnormal division events. Cell and nuclear morphology were both altered in confinement, with the most confining channels preventing cells from undergoing the normal increase in size from G1 to S/G2/M during cell cycle progression. Finally, our results suggest that confinement induces a mechanical memory to the cells, given our observation of lasting effects on cell division and morphology, even after cells exited confinement. Together, our results provide new insights into the possible impact of mechanical forces on primary and secondary tumor formation and growth.

Introduction

Cell growth and behavior have been shown to be dictated not only by soluble and matrix-bound biochemical cues, but also by the physical microenvironment, which includes factors such as matrix rigidity, shear stress, topography, and confinement [1,2]. Initial studies sought to understand primarily the first three cues, but technical advances in microfabrication techniques and biomaterial synthesis have enabled insights into the impact of physical confinement on cell behavior within the last decade. Tumor cells experience confinement within the primary tumor, due to cell crowding caused by increased growth rates of cancer cells, and during metastasis, as tumor cells circulate through the blood stream within capillaries as small as 5 μm in diameter, intravasate or extravasate into or out of blood vessels, and migrate along anatomic features such as muscle and nerve fibers [3,4]. Intriguingly, molecular mechanisms driving cell behavior can be altered in response to physical confinement. For example, some cell types migrating through three-dimensional physical confinement primarily use mechanisms driven by acto-myosin contractility [5–8], but other cell types can employ a migration mechanism based on water permeation through aquaporins [9], a nuclear piston-based mechanism for migration in a 3D matrix [10], a mechanism based on asymmetric hydraulic pressure where cells can “push” water during migration in microchannels [11], and a mechanism whereby the cell’s nucleus ruptures to promote squeezing through confined spaces [12].

Within confinement, the nucleus has been demonstrated to be the rate limiting step for cell migration [13,14], where altered nuclear shape and structure might also cause altered nuclear function. Indeed, the nucleus is responsible for regulating cell growth, replication, and ultimately division through various signaling cascades [15–21]. Nuclear shape throughout cell division is an important regulator of cell cycle progression. The cell and nucleus enlarge during G1 to provide adequate room for DNA synthesis to occur in the nucleus during S stage, and the nucleus prepares for mitosis during G2, where the nucleus controls chromatid formation and correct alignment for daughter cell formation [22]. The main cytoskeletal elements involved in cell division are microtubules, which control mitotic spindle formation within the nucleus and pull respective chromatin into separate areas of the cell to create identical daughter cells, and cortical actin, which drives cell rounding. During cell division, cell rounding is an important step to allow for proper spindle assembly and positioning and subsequently, correct daughter cell formation [22–24].

Given the critical role of nuclear shape and mechanics in cell division, we hypothesized that physical confinement of the cell body, and also the cell’s nucleus, would alter cell cycle progression. Indeed, recent reports have shown that confinement in a 3D spheroid alters nucleus geometry and delays cell division [25], and that imposing confining forces on cells in the vertical direction impairs mitotic spindle rounding, results in abnormal daughter geometries, and can promote division into more than two progeny [26,27]. Meanwhile, the effects of bi-axial cell confinement should also be investigated, since intravital microscopy recently revealed that metastatic tumor cells migrate through small (1–30 μm) spaces in vivo, including through preexisting longitudinal microtracks in tissues [28,29]. Hence, we sought to understand the effects of bi-axial physical confinement on cell cycle progression and ultimately, division, using microfabricated devices with microchannels of defined geometry and varying cross-sectional area. Using a sarcoma cell line stably transfected with the fluorescence ubiquitination cell cycle indictor (FUCCI), we could visually distinguish cells between G1 and S/G2/M stages of the cell cycle and found that in the most confining microchannels, progression through the S/G2/M stage (but not G1) was delayed or halted. Together, our results suggest that as tumor cells migrate through physically confining spaces, cell division may either (1) be halted until the cell enters a less restrictive space, or (2) proceed but result in more frequent instances of abnormal division events that could lead to secondary tumor formation.

Results

Physical confinement decreases sarcoma cell divisions and restricts cleavage furrow angle

Given the physiological role of physical confinement in multiple stages of tumor progression, we first evaluated the effects of bi-axial confinement on the fraction of sarcoma cells that divided over the course of 18 hours, which is significantly longer than the 13-hour cell cycle length of these cells. We used two assays: (1) cells seeded on fibronectin-coated glass-bottom dishes, which served as a two-dimensional (2D) control (Figure 1(a)), and (2) cells seeded in a microfluidic device containing fibronectin-coated microchannels of 10 µm height and varying width (3, 6, 10, 20, and 50 µm) (Figure 1(b)). A chemotactic gradient was initially set up within the microfluidic device to promote cell migration into the microchannels. After 24 hours, the gradient was removed and replaced with media containing full serum (Figure 1(b)) to reduce cell migration persistence and increase cells’ time spent in channels. Cells divided within the microchannels of all widths, as demonstrated by imaging via phase contrast microscopy, with the morphology of cells during division in the widest 50 µm channels resembling that on 2D surfaces (Figure 1(c)). However, the fraction of cells that divided within all microchannel widths was significantly reduced in comparison with the 2D substrates (Figure 1(d) and Supplementary Table S1). In addition, the fraction of cells dividing in higher degrees of confinement (3 μm wide channels) was significantly lower in comparison with wider 10, 20, 50 μm channels (Figure 1(d) and Supplementary Table S1).

Figure 1.

Cell divisions decrease in confinement. Schematic of (a) 2D controls and (b) microchannel confinement devices. (c) Representative phase contrast images of cells dividing on 2D substrates, as well in 50, 20, 10, 6, and 3 μm wide microchannels. Scale bar for image showing 2D control represents 10 μm. (d) Effects of channel width on fraction of cells that divided, including normal and abnormal divisions. Bars that do not share a letter are statistically different via ANOVA with post-hoc Tukey test. (e) Breakdown of cell division events in all channel widths. (f) Representative image of quantification method for angle of cell division in 50 and 3 μm wide devices, where x-axis is parallel to axis of channel. (g) Effect of channel width on angle of cell divisions. Each point represents one cell. (h) Gaussian curve fits to histograms of angle of division. Cell counts and full statistical comparisons for panel D are displayed in Supplementary Table S1. Error bars represent standard error (d) and standard deviation (g).

Taking into account the abnormal cell divisions previously reported in cells that were confined in the z-direction [26], we broke down the division events in confinement. Cells in the 3 and 6 μm wide channels experienced more abnormal division events than cells in the 10, 20, and 50 μm wide channels (Figure 1(e)). Cells in the 3 and 6 μm wide channels showed evidence of a distinct cleavage furrow and two individual daughter cells, but in some cells, the two daughter cells fused together to create one multi-nucleated cell in the G1 phase (Supplementary Movies 1 and 2). Furthermore, both 3 and 6 μm wide channels contained cells dividing with a distinct cleavage furrow but with two daughter cells that immediately died upon completion of cytokinesis in the channels. In addition, some cells in the 6 μm wide channels divided into 3 individual daughter cells (Supplementary Movie 3). These are all in comparison to cells in the 10, 20, and 50 μm wide channels that displayed a majority of normal divisions (Supplementary Movie 4) and very few abnormal divisions (Figure 1(e)). Increasing confinement also restricted the angle of division, where cells in 3 and 6 μm wide channels were forced to divide with a cleavage furrow angle at 90 degrees with respect to the length of the microchannel, while cells in wider channels were free to divide in any direction (Figure 1(f)). Indeed, quantification of these results showed a significantly larger spread of cleavage furrow angles on 2D substrates and in wider microchannels, while cleavage furrow angles in narrower channels were limited to around 90 degrees (Figure 1(g)). Histograms demonstrated a relatively even distribution of all cleavage furrow angles in cells dividing in 2D control experiments, and cleavage furrow angle distributions centered about 90 degrees for cells dividing in microchannels, with the distribution width decreasing with increasing confinement (Figure 1(h)).

Physical confinement alters cell and nuclear morphology during the cell cycle

We sought to determine why confinement reduced the fraction of sarcoma cells that divided, and we hypothesized that cell and/or nuclear morphology and deformation were playing a role. Taking advantage of the FUCCI system that allows for visual distinction of cells in the G1 versus S/G2/M stage (Figure 2(a)), we further evaluated the effects of confinement on sarcoma cell division. First, we confirmed that cells indeed expressed Cdt1-RFP during the G1 phase of the cell cycle, fluoresced yellow during transition from G1 to S/G2/M, expressed geminin-GFP during the S/G2/M phase, and turned colorless while undergoing mitosis (Figure 2(a)). Using phase contrast images in conjunction with the FUCCI images, we outlined both the cells and their nuclei (Figure 2(b)) and quantified morphology parameters of each, as described in the Materials and Methods section. Cell area (Figure 2(c)) decreased in all degrees of confinement as compared to cells in 2D unconfined controls (Figure 2(d) and Supplementary Table S2). The area of G1 stage cells was mostly similar across channel widths, while the area of S/G2/M stage cells decreased with increasing confinement (Figure 2(d) and Supplementary Table S2). Interestingly, cells in the G1 phase had smaller areas than those in the S/G2/M stage, except for the case of the 3 μm wide channels, where the trend was opposite (Figure 2(d) and Supplementary Table S2). We expected cells in the S/G2/M stage to have replicated their DNA content and also enlarged to prepare for mitosis, which is consistent with our results in channels ≥ 6 μm wide, but the data also suggested that the highest degrees of bi-axial confinement (i.e. 3 μm wide channels) could have a growth limiting effect on cells preparing for division. Nuclear area (red fluorescence in Figure 2(c)) in the G1 stage was not significantly different across the range of channel widths (Figure 2(e) and Supplementary Table S2). Meanwhile, nuclear area in the S/G2/M stage displayed a biphasic trend across channel widths (Figure 2(e) and Supplementary Table S2). Similar to cell area, cells in the G1 phase had smaller nuclear areas than those in the S/G2/M stage, except for the case of the 3 μm wide channels, where the trend was opposite (Figure 2(e) and Supplementary Table S2).

Figure 2.

Cell and nuclear morphologies are altered in confinement. (a) Schematic depicting FUCCI reporter construct design and corresponding fluorescent expressions to cell cycle stages. Scale bar for images showing FUCCI progression represents 10 μm. (b) Representative images of cell morphology in various channel widths. Fluorescent images showing FUCCI expression were overlaid on the phase contrast images, and white dotted lines were added to improve visualization of cell outlines. (c) Image demonstrating area calculation. White dotted outline represents manual tracing of cell border, while red fluorescence from FUCCI expression was used to calculate nuclear area. (d) Cell area and (e) nuclear area as a function of channel width. (f) Image demonstrating aspect ratio calculation. (g) Cell aspect ratio and (h) nuclear aspect ratio as function of channel width. (i) Image demonstrating determination of nuclear major and minor length. (j) Major and minor axes of nuclei in G1 stage and (k) S/G2/M stage as a function of channel width. Cell counts and full statistical comparisons are displayed in Supplementary Table S2 (for panels d, e, g, and h) and S3 (for panels j and k). Error bars represent standard error.

We also quantified the cellular aspect ratio (Figure 2(f)), where an aspect ratio of 1 correlated with a highly circular cell, and aspect ratios > 1 tended towards more elongated cells. Most notably, the aspect ratio of cells in both G1 and S/G2/M increased as channel width decreased (Figure 2(g) and Supplementary Table S2). Nuclear aspect ratio trends also matched those of the cells themselves, where nuclear aspect ratio increased with decreasing channel width (Figure 2(h) and Supplementary Table S2). Breaking down nuclear morphology further, we compared the major and minor axis lengths of the nucleus fit to an ellipse (Figure 2(i)). In both the G1 and S/G2/M stages, the major axis of the nuclear ellipse decreased as channel width increased, while the minor axis increased as channel width increased (Figure 2(j,k) and Supplementary Table S3). It seems that once the nucleus becomes confined (3 and 6 μm channels; Figure 2(b)) the cell experiences a growth constraining force that impacts the S/G2/M stage.

Physical confinement alters sarcoma cell migration during the cell cycle

Next, we aimed to understand if migration through confinement could be playing a role in the reduced division frequency in confinement, possibly through alteration of time spent in the channels. Previous studies by us and others have shown that tumor cells display novel mechanisms of migration through confined spaces [5–12], and that there are cell cycle stage dependent differences along 3D collagen fibers [30]. Using a combination of phase contrast and fluorescence timelapse imaging, we tracked cells migrating through the microchannels in our fabricated devices (Figure 3(a)), and then quantified cell speed (Figure 3(b)) and persistence (Figure 3(c)). For cells in the G1 phase, cell speed in all microchannel widths was significantly greater than cell speed in 2D controls (Figure 3(b) and Supplementary Table S4). For cells in the S/G2/M phase, cell speed in wider channels (10, 20, 50 μm) was significantly larger than cell speed in 2D controls, while cell speed in narrow channels (3, 6 μm) was not significantly different from 2D controls (Figure 3(b) and Supplementary Table S4). Cell speed in G1 and S/G2/M phases generally decreased with increasing confinement, though there was a slight statistically significant biphasic behavior in the G1 phase across channel widths (Figure 3(b) and Supplementary Table S4). Interestingly, cells in 2D controls had similar speeds in G1 and S/G2/M phases, while all cells in all microchannel widths were significantly faster in the G1 phase in comparison with the S/G2/M phase (Figure 3(b) and Supplementary Table S4). In both G1 and S/G2/M phases, the persistence of migrating cells increased with confinement from 2D controls to 10 μm wide microchannels and then remained the same between 10, 6, and 3 μm wide channels (Figure 3(c) and Supplementary Table S4). Hence, while migration speed was dependent on microchannel width, reduced cell division in narrower channels cannot be attributed to the cells spending less time in the channels, since the cells did not migrate faster in these channels.

Figure 3.

Cell migration is altered in confinement. (a) Representative images of cells in G1 and S/G2/M stages migrating through 6 μm wide channels. (b) Cell speed and (c) persistence as a function of channel width. Cell counts and full statistical comparisons for panels B and C are displayed in Supplementary Table S4. Error bars represent standard error.

Confinement increases time spent in the S/G2/M phase of the cell cycle

Considering the results from the morphological and migration analysis, we hypothesized that physical confinement (i.e. 3 and 6 μm wide channels) impacts the S/G2/M stage of the cell cycle, possibly by preventing cell division or lengthening time spent in S/G2/M. To probe this idea further, we compared the time that cells spent in G1 versus S/G2/M stages of the cell cycle in the varying channel widths. In line with our hypothesis, the time that cells spent in the G1 stage of the cell cycle was independent of degree of confinement (Figure 4(a) and Supplementary Table S5). However, the time that cells spent in the S/G2/M stage drastically increased as channel width decreased (Figure 4(a) and Supplementary Table S5). Cells in the 3 and 6 μm wide channels spent more than double the time in the S/G2/M stage than cells in 10, 20, 50 μm wide channels and cells in 2D controls, with most cells still in the S/G2/M phase by the end of the timelapse experiment. Indeed, within the microchannels, fraction of cell division (from Figure 1(d)) seemed to correlate with the average time cells spent in channels (Figure 4(b)). In line with this, the cumulative fraction of cells that successfully transitioned from the G1 to S/G2/M phase was similar across channel widths (Figure 4(c)), while the cumulative fraction of cells that divided decreased with decreasing channel width (Figure 4(d)). Hence, time spent in the G1 phase remained unaffected by confinement, while the S/G2/M stage was primarily affected, thus supporting our hypothesis that confinement decreases the fraction of cells that divide by altering the S/G2/M stage of the cell cycle.

Figure 4.

Confinement lengthens time spent in S/G2/M, but not G1, stage of cell cycle. (a) The amount of time that cells spend in G1 stage in confinement as a function of channel width. Upward pointing arrows above bars for 3 and 6 μm wide channels for S/G2/M stage indicate that the actual values extend longer than the length of the timelapse. The values for the bars shown were calculated using the last frame of the time lapse as the end of the S/G2/M phase and therefore underestimate the actual length of the S/G2/M phase. (b) The fraction of cells that divide (also shown in Figure 1(d)) versus the average amount of time that cells spent in various channel widths. Data point labels indicate microchannel width. (c) Cumulative fraction of cells that progress from G1 stage to S/G2/M stage over time for varying channel widths. Only cells that transitioned from G1 to S/G2/M over the course of the timelapse were counted in the fraction denominator. (d) Cumulative fraction of cells that divide over time for varying channel widths. Only cells that entered the channel over the course of the timelapse were counted in the fraction denominator. Cell counts and full statistical comparisons for panel A are displayed in Supporting Table S5. Error bars represent standard error.

Some effects of confinement disappear upon cell exit from confinement

We sought to understand the degree to which confinement affected cells after they had migrated through the microchannels and returned to an unconfined environment. Therefore, we tracked and analyzed cells in the upper region of the device and compared them to cells in the microchannels (Figure 5(a)). We found that cells post-confinement divided less frequently than cells in 2D unconfined controls, but more frequently than cells in confinement (Figure 5(b) and Supplementary Table S6). Cells post-confinement spent similar amounts of time in each stage of the cell cycle as compared to cells in 2D controls (Figure 5(c) and Supplementary Table S7), suggesting that the effects of high degrees of confinement in lengthening the S/G2/M phase of the cell cycle were reversed upon exit from confinement. Cells post-confinement were larger than cells in confinement for all microchannel widths and for both stages of the cell cycle but were still smaller on average than cells in the 2D controls for the corresponding cell cycle stage, which could contribute to the reduced divisions post-confinement (Figure 5(d) and Supplementary Table S8). Interestingly, nuclear area of cells post-confinement was smaller than that of cells in confinement, and also significantly smaller than 2D controls for the S/G2/M stage cells (Figure 5(e) and Supplementary Table S8). Finally, the cell (Figure 5(f)) and nuclear (Figure 5(g)) aspect ratios, and the cell speed (Figure 5(h)) and persistence (Figure 5(i)) mostly were similar to 2D unconfined control values when the cells were post-confinement (Supplementary Tables S8 and S9). Based on these results, it seems that confinement has a transient effect on cell migration and the time regulation of the cell cycle stages. However, traveling through confinement may have a lasting effect on cellular ability to divide and morphological area, at least over a 16 hour time window.

Figure 5.

Some effects of confinement disappear upon cell exit from confinement. (a) Representative schematic of cells in confinement versus cells post-confinement. Cells were observed post-confinement for up to 16 hours. Scale bar on post-confinement image represents 20 μm. (b) Comparison of fraction of cells dividing as a function of channel width during and post confinement. (c) Time spent in G1 and S/G2/M stages of the cell cycle as a function of channel width, during and post confinement. (d) Cell area and (e) nuclear area as a function of channel width during and post confinement. (f) Cell aspect ratio and (g) nuclear aspect ratio as a function of channel width during and post confinement. (h) Cell speed and (i) cell persistence as a function of channel width during and post confinement. Cell counts and full statistical comparisons are displayed in Supporting Tables S6 (for panel b), S7 (for panel c), S8 (for panels d-g), and S9 (for panels h-i). Error bars represent standard error.

Discussion

There has been a growing appreciation in the field of cancer metastasis and cell mechanobiology for the role of physical confinement in many aspects of tumor cell behavior, including cell division. Furthermore, it is becoming evident that tumor cells must navigate heterogenous microenvironments during metastasis, including microchannels that impose bi-axial physical confinement on the cell body and nucleus [28,29]. Hence, in this work, we sought to understand how systematic control of bi-axial confinement in a microchannel device impacts cancer cell cycle progression. Our work, using cells derived from mouse sarcoma tissue in combination with the FUCCI reporter system, indicates that bi-axial physical confinement (1) reduces frequency of cell division; (2) increases frequency of abnormal division events; (3) constricts the cleavage furrow angle; (3) alters cell and nuclear morphology, with the most confining channels preventing cells from increasing in size in the x-y plane from G1 to S/G2/M during cell cycle progression; (4) alters cell migration speed and persistence; (5) increases time spent in the S/G2/M phase of the cell cycle and/or halts progression through mitosis, and (6) has some lasting effects on cell division and morphology, even after cells have exited confinement.

It has been established for several decades that cancer cell growth occurs mainly on the peripheral outer layers of the tumor [31]. Circulating tumor cells have the ability to become mitotic during late stage aggressive cancers [32], and metastatic tumor cells in the blood stream have been shown to attach to the endothelium and proliferate within the vasculature [33]. Following extravasation, cancer cells attach to the extravascular side of vessels, flattening themselves out and enlarging, then dividing rapidly [34]. This literature is in line with our results that less confined cells (outer layers of the tumor) experience more cell division events than those in the interior (higher degrees of confinement). Furthermore, the increased abnormal division events occurring in confinement may be associated with chromosomal abnormalities that are a hallmark of cancer. It is also worthwhile to note that single multinucleated sarcoma cells have been previously shown to form solid tumors [35].

Recent reports have shown that cells undergoing mitosis in 3D collagen matrices maintain a highly elongated shape, and this behavior can be recapitulated in collagen-coated microchannels independent of β1 integrin [36,37]. However, these studies have been limited in their inability to distinguish other stages of the cell cycle or assess beyond the cleavage furrow. Our work supplements these findings, as our primary focus was not M stage but rather the cell cycle as a whole, by demonstrating that confinement imposes morphological and migratory abnormalities at potentially earlier stages than just mitosis. It has also been shown that external forces and uni-axial confinement prevent actin cortex-driven cell rounding, which can lead to incorrect spindle assembly, pole splitting, and delays in mitotic progression [26]. Furthermore, in some tumor cells, microtubule inhibition results in drastic migratory changes in cells in 3D physically confining microenvironments, as compared to perturbations of integrins, acto-myosin, or the Rho/ROCK pathway [26,38,39]. Indeed, our previous work has demonstrated that bi-axial confinement promotes a more diffuse actin and microtubule network in cells [9,38,40], and therefore it is possible that these altered cytoskeletal arrangements may improperly position the mitotic spindle, thus preventing division. Along with this, uni-axial confinement has been reported to increase the number of abnormal division events and delay mitotic progression caused by the incorrect spindle machinery assembly. Our work supports the appearance of these abnormal division events in bi-axial confinement as well.

We also found it interesting that the highest degrees of physical confinement resulted in deformation of the nucleus (i.e. in 3 and 6 µm channels), as well as a lack of cell area increase from G1 to S/G2/M (i.e. in 3 µm channels), as compared to wider channels. Nuclear deformation due to mechanical forces can lead to numerous alterations in cells, including restructuring of the lamin A/C network, increased mobility of chromatin, redistribution of biophysical stresses within the cell, as well as time-dependent transcriptional silencing or activation (reviewed in [41]). These changes reflect a possible mechanism by which confinement lengthens the later stages of the cell cycle and promotes abnormal division events. It is also possible that there exist cell cycle checkpoints that are tension-dependent. Furthermore, the unexpected decrease in x-y projected cell area from G1 to S/G2/M in 3 µm channels leads us to hypothesize that confinement promotes cell volume dysregulation, thus preventing cells from undergoing the normal mitosis-associated increase in cell volume while cell material is replicating.

We have previously shown through experimental and theoretical analysis that tumor cells can utilize aquaporin water channels for migratory mechanisms in confinement [9]. Similarly, recent theoretical work has shown that active pumping of water and ions at the cell poles can explain cell shape changes during cytokinesis, and that this process can be actomyosin-independent assuming that the intercellular fluid flow is above a threshold magnitude [42]. However, they also suggest it is likely that water flux works together with the actomyosin cortex ring and actin cortex to regulate cell shape during division [42]. We are currently working to understand the effects of aquaporin distribution and water flux on tumor cell cycle progression and division in confinement, given the striking connection between cell shape, migration, and confinement.

The tumor microenvironment is not spatially or temporally homogeneous, as cells are constantly exposed to varying environments as they move and proliferate. Interestingly, non-neural cells (especially stem cells) display mechanical memory of their exposure to stiffness or fluid flow, and these memory signals can be transmitted bioelectrically, via miRNA, or through the YAP/TAZ pathway [43–45]. There can also be epigenetic regulation of bivalent chromatin domains on cancer cells in response to drug re-expression [46]. In this regard, our results suggest that cells may retain a mechanical memory of physical confinement, thus affecting cell size and ability to divide even after cells leave confinement. Future work could elucidate if this is truly mechanical memory and how, whether epigenetically, or genetically, this memory is imposed, and whether it is temporal or permanent.

Our combination of FUCCI-transfected sarcoma cells and microchannel devices presents several advantages over other in vitro systems used to study cell division. First, the device design and fabrication procedure allow us to impose systematic control of bi-axial confinement on cells. Second, the device provides excellent imaging capabilities, both in phase contrast and fluorescence microscopy, given that the bottom surface is a glass coverslip. Hence, we could quantitatively analyze and compare multiple parameters of cell division, cell cycle progression, cell morphology, and cell migration in varying degrees of physical confinement. Third, using a cell line stably transfected with the FUCCI vector allowed us to circumnavigate other hurdles associated with transient transfection, such as loss of FUCCI expression over the course of the experiments, low transfection efficiency, and possibility for cells to be expressing Cdt1-RFP but not geminin-GFP, or vice versa. We encountered these issues when attempting to create new cell lines with Fisher Scientific’s Premo FUCCI Cell Cycle Sensor and BacMam 2.0 delivery system.

We also acknowledge several limitations of our work. First, the microchannel devices were composed of PDMS, which has a stiffness in the MPa range, larger than most physiological tissues which are in the kPa range. However, we note that our goal in this study was to explore the effects of bi-axial confinement, not stiffness, on cell cycle progression. Other labs have recently developed other systems, such as polyacrylamide-based devices, to alter microenvironment stiffness [6], and hence future work could explore the interplay between confinement and stiffness on cell cycle progression. Second, this work was carried out on only one cell line that was stably transfected with FUCCI, but ideally we would have used multiple cell lines to determine whether our results are cell line-dependent or general phenomena. As mentioned above, we attempted to transfect several other cell lines, including MDA-MB-231 and human bone marrow-derived mesenchymal stem cells, with Fisher Scientific’s Premo FUCCI Cell Cycle Sensor using the BacMam 2.0 delivery system. However, we experienced extremely low transfection rates of both Cdt1-RFP and geminin-GFP, which would have prevented us from gathering sufficient numbers of cells in the microchannel devices to form meaningful conclusions on those cell lines. Hence, our future work will be aimed at using other FUCCI vectors and delivery systems to create new stable cell lines expressing FUCCI.

Other future work should focus on exploring whether there are specific molecular signaling pathways or processes that prolong the S/G2/M phase in confinement, and whether cell cycle checkpoints are affected, perhaps in a tension-dependent way. We note that we did perform some experiments in which cells were treated with indisulam and RO-3306 (data not shown), previously shown to induce cell cycle arrest in G1 and G2, respectively. On 2D fibronectin coated plates, we observed a near complete cell cycle arrest in the corresponding cell cycle stage. Cells treated with indisulam were morphologically identical to untreated cells in the G1 stage; meanwhile, cells treated with RO-3306 were noticeably larger and more circular than untreated cells in the S/G2/M stage. However, we experienced difficulties seeding the cells into our PDMS microfluidic devices and retaining cell cycle inhibition, which may be due to the absorption of the drug by the surrounding PDMS [47].

In summary, we have integrated mouse sarcoma cells stably expressed FUCCI into microfluidic devices that impose bi-axial physical confinement during cell migration in microchannels and shown that confinement reduces frequency of cell division while increasing frequency of abnormal division events, which in other work has been shown to lead to solid tumor formation. Confinement also alters cell and nuclear morphology, with the most confining channels preventing cells from increasing in size from G1 to S/G2/M during cell cycle progression, and with lasting effects even after exit from confinement. Finally, confinement does not seem to affect the G1 phase of the cell cycle, but increases time spent in the S/G2/M phase of the cell cycle and/or halts progression through mitosis. Together, our results suggest that as tumor cells migrate through physically confining spaces, cell division may either (1) be halted until the cell enters a less restrictive space, or (2) proceed but result in more frequent instances of abnormal division events that could lead to secondary tumor formation. Finally, our results provide new insights into the possible impact of mechanical forces on primary and secondary tumor formation and growth.

Methods

Cell culture

S180 mouse sarcoma cells stably transfected with a fluorescent ubiquitin cell cycle indicator (FUCCI) were gifted from Jean Paul Thiery, who is currently at the Cancer Science Institute of Singapore. Cells expressed Cdt1-RFP during the G1 phase of the cell cycle and geminin-GFP during the S/G2/M phase of the cell cycle (Figure 2(a)). Cells were grown in medium comprised of Dulbecco’s modified Eagle’s medium with high glucose (ThermoFisher Scientific, https://www.thermofisher.com/order/catalog/product/11965092) supplemented with 10% heat inactivated fetal bovine serum (ThermoFisher Scientific, https://www.thermofisher.com/order/catalog/product/10438026?SID=srch-srp-10438026), and 1% penicillin/streptomycin 1000 U/mL (P/S; ThermoFisher Scientific, https://www.thermofisher.com/order/catalog/product/15140122?SID=srch-srp-15140122). Cells were passaged when they reached approximately 90% confluency and passages up to 30 were used in experiments. Cells were washed with phosphate buffered saline (PBS; VWR, https://us.vwr.com/store/product/10243646/phosphate-buffered-saline-pbs-corning) and lifted using 0.25% Trypsin-EDTA (ThermoFisher Scientific, https://www.thermofisher.com/order/catalog/product/25200056?SID=srch-srp-25200056). All cells were cultured at 37C, 50% humidity, and 5% CO2: 95% air.

Microfluidic device fabrication

Microfluidic devices were fabricated as previously described [38,48]. All fabrication was carried out in the University of Maryland Nanocenter Fabrication Lab. To summarize, two masks were designed in AutoCad (AutoDesk), one with the microchannels, the other with the larger feed lines. A layer of SU-8 3010 negative photoresist (MicroChem, http://www.microchem.com/Prod-SU83000.htm) was spincoated onto a silicon wafer (University Wafer, https://order.universitywafer.com/default.aspx?cat=Silicon). An EVG620 Mask Aligner (EV Group) was used to UV crosslink the SU-8 through the mask containing the microchannel design. The uncrosslinked photoresist was washed away using the SU-8 developer (MicroChem). A second layer of SU-8 3025 negative photoresist was spincoated on the wafer, and the mask containing the feed lines was aligned and the photoresist was UV crosslinked again. Excess SU-8 3025 was dissolved and washed away. Wafers were then silanized using tridecafluoro-1,1,2,2,tetrahydrooctyl-1-trichlorosilane (97%; Pfaltz & Bauer, https://us.vwr.com/store/product/21748948/tridecafluoro-1-1-2-2-tetrahydrooctyl-1-trichlorosilane-97) overnight in a vacuum desiccator. The completed silicon wafer contained microfluidic devices with microchannels of 3, 6, 10, 20, 50 μm in width; all channels were 200 μm in length and 10 μm in height, and the feed lines were 50 μm in height. Polydimethylsiloxane (PDMS, Robert McKeown Company, http://products.robertmckeown.com/item/potting-encapsulants/silicone-potting-encapsulants/sylgard-174-184-silicone-elastomer) was mixed at a 10:1 base: crosslinker ratio, poured over the silicon wafer, vacuum desiccated for 1 hour, and baked at 85°C for 3 hours. The PDMS was removed from the wafer, cleaned with ethanol and water, dried at 85°C for 5 minutes and then plasma treated for 2.5 min. The PDMS devices were then bonded to glass coverslips (which were also cleaned and plasma treated simultaneously with the PDMS) for 5 minutes and the PDMS device-coverslip unit was UV sterilized for 10 minutes. 20 ug/mL fibronectin (Sigma-Aldrich, https://www.sigmaaldrich.com/catalog/product/sigma/f2006?lang=en®ion=US) was added to all wells of the device and allowed to adsorb for one hour. The device was then washed with PBS 3 times for 5 minutes each to remove excess protein. Cells were added according to the description in the section below.

Microscopy time lapse experiments

Cells were washed with PBS and then lifted using 0.25% trypsin-EDTA as previously mentioned. Cells were counted using a hemocytometer and resuspended in DMEM containing 1% P/S to give a final concentration of 4 × 10⁵ cells/mL. Twenty five μL of the final cell suspension was added to the inlet of the lower feed line and allowed to incubate in the device for 5 minutes. Excess media was removed from the lower feed line inlet and outlet lines and DMEM containing 1% penicillin/streptomycin 1000 U/mL was added to all wells, except for the topmost upper feed line well, which received DMEM containing 1% P/S and 10% FBS to serve as the chemoattractant to encourage cells to migrate through the microchannels. The device was the cultured for 24 hours and then all media was replaced with DMEM containing 1% P/S and 10% FBS in all wells. The device was then imaged on an Olympus IX83 microscope (Olympus) using a 20× objective. For 2D control experiments, we used 24 well glass bottom dishes (Mattek, https://www.mattek.com/store/p24g-1-5-13-f-case/) that were coated with 20 ug/mL fibronectin which was allowed to adsorb for 1 hour, and subsequently they were washed 3 times with PBS for 5 minutes each. Cells were plated on the glass at a concentration of 2 × 10⁵ cells/mL and allowed to attach and grow for 24 hours before they were imaged on the Olympus IX83 microscope. A chamber adjusted to 37°C, 50% humidity, and 5% CO₂:95% air was used on the microscope stage to sustain cell viability. Images were taken at 10 min intervals for durations longer than 18 hours. Following overnight timelapse, cells plated on 24 well glass bottom plates were imaged using a 20x objective for morphological analysis.

Data & statistical analysis

All cell and nuclear morphological parameters were manually traced and quantified using ImageJ. Cell and nuclear perimeters were traced to quantify area and aspect ratio (ratio of the width to the length). The outline of the nucleus was fit to an ellipse and the major and minor axes were calculated. Cells were tracked using the Manual Tracking ImageJ plug in and then speed and persistence values were calculated using a custom written Matlab code by KMS. We calculated persistence as the end to end distance the cell traveled divided by the total distance. Data for cells was pooled from at least 3 independent trials and assumed to be normally distributed. An ANOVA with post-hoc Tukey test was conducted to evaluate significance across all samples, as presented in the Supplementary Tables S1-S9. A significance level of p = 0.05 was used, and error bars represent SEM unless otherwise noted in the figure caption.

Funding Statement

This work was supported by the Burroughs Wellcome Fund.

Acknowlegments

We thank Jean Paul Thiery for generously providing S180 cells stably transfected with the FUCCI reporter (i.e. AB3 cells). We acknowledge the support of the Maryland NanoCenter and its FabLab for providing photolithography resources. This work was supported by a Burroughs Wellcome Career Award at the Scientific Interface (to KMS), the Fischell Department of Bioengineering, the University of Maryland, and a Summer Research Fellowship from the University of Maryland Graduate School (to RAM).

Disclosure statement

No potential conflict of interest was reported by the authors.

Supplementary material

Supplementary data for this article can be accessed here.