Molecular Insights into Division of Single Human Cancer Cells in On-Chip Transparent Microtubes

In vivo, mammalian cells proliferate within 3D environments consisting of numerous microcavities and channels, which contain a variety of chemical and physical cues. External environments often differ between normal and pathological states, such as the unique spatial constraints that metastasizing cancer cells experience as they circulate the vasculature through arterioles and narrow capillaries, where they can divide and acquire elongated cylindrical shapes. While metastatic tumors cause most cancer deaths, factors impacting early cancer cell proliferation inside the vasculature and those that can promote the formation of secondary tumors remain largely unknown. Prior studies investigating confined mitosis have mainly used 2D cell culture systems. Here, we mimic aspects of metastasizing tumor cells dividing inside blood capillaries by investigating single-cell divisions of living human cancer cells, trapped inside 3D rolled-up, transparent nanomembranes. We assess the molecular effects of tubular confinement on key mitotic features, using optical high- and super-resolution microscopy. Our experiments show that tubular confinement affects the morphology and dynamics of the mitotic spindle, chromosome arrangements, and the organization of the cell cortex. Moreover, we reveal that membrane blebbing and/or associated processes act as a potential genome-safety mechanism, limiting the extent of genomic instability caused by mitosis in confined circumstances, especially in tubular 3D microenvironments. Collectively, our study demonstrates the potential of rolled-up nanomembranes for gaining molecular insights into key cellular events occurring in tubular 3D microenvironments in vivo.


* S Supporting Information
ABSTRACT: In vivo, mammalian cells proliferate within 3D environments consisting of numerous microcavities and channels, which contain a variety of chemical and physical cues. External environments often differ between normal and pathological states, such as the unique spatial constraints that metastasizing cancer cells experience as they circulate the vasculature through arterioles and narrow capillaries, where they can divide and acquire elongated cylindrical shapes. While metastatic tumors cause most cancer deaths, factors impacting early cancer cell proliferation inside the vasculature and those that can promote the formation of secondary tumors remain largely unknown. Prior studies investigating confined mitosis have mainly used 2D cell culture systems. Here, we mimic aspects of metastasizing tumor cells dividing inside blood capillaries by investigating single-cell divisions of living human cancer cells, trapped inside 3D rolled-up, transparent nanomembranes. We assess the molecular effects of tubular confinement on key mitotic features, using optical high-and super-resolution microscopy. Our experiments show that tubular confinement affects the morphology and dynamics of the mitotic spindle, chromosome arrangements, and the organization of the cell cortex. Moreover, we reveal that membrane blebbing and/or associated processes act as a potential genome-safety mechanism, limiting the extent of genomic instability caused by mitosis in confined circumstances, especially in tubular 3D microenvironments. Collectively, our study demonstrates the potential of rolled-up nanomembranes for gaining molecular insights into key cellular events occurring in tubular 3D microenvironments in vivo. KEYWORDS: rolled-up nanofilms/membranes, mitosis, chromosome segregation, membrane blebbing, actin cortex, 3D cell culture, metastasis C ancer cells, originating from primary lesions can invade adjacent tissues and squeeze along muscle fibers, 1 before entering the vascular system and traveling to distal sites, where some of them can proliferate and form metastases. During this process cancer cells can become trapped inside narrow blood capillaries 2−5 ( Figure 1A), where they adapt to the tubular confinement by elongating into cylindrical shapes 6 (Figure 1A and B). Importantly, several in vivo studies have shown that circulating tumor cells can divide under such tubular confinements and form early micrometastatic colonies before exiting the vasculature. 2,3 This intravascular proliferation in tubular environments is thought to be a critical step during metastasis, at least in the lung and the liver. 2,3,7−9 Despite the leading role of metastasis in cancer mortality, many steps during this process remain poorly understood. This is due to metastasis being a rarebut often destructiveprocess that happens in parts of the body not easily accessible to molecular analyses, such as the vascular or ductal system. Gaining molecular insights into how 3D, and in particular tubular, constraints affect the division of human cancer cells might therefore lead to a better understanding of the events and factors that promote metastasis, which in the future could help improve antimetastatic cancer treatments.
Cell division, the partitioning of the nucleus (mitosis) followed by the division of the cytoplasm (cytokinesis), involves striking 3D changes in the cell's geometry and cytoskeleton. For instance, most dividing eukaryotic cells restructure their interphase actin structures, 10 resulting in the recruitment of actin filaments to the cell cortex, 11 thereby imparting a heightened stiffness to mitotic cells. 12,13 This increased rigidity is usually accompanied by conspicuous changes in mitotic cell shape, whereby the cells abandon their flattened and spread-out interphase morphologies on planar substrates to take up rounded-up sphere-like shapes. 14,15 The cell cortex is then further remodeled to generate a contractile Figure 1. Tubular confinement affects cell shape, chromosome alignment, and spindle morphology in mitotic HeLa cells. (A) Schematic of metastasizing tumor cells originating from the primary tumor and circulating the blood system. The large tumor cells get trapped, for instance, through size restriction, in narrow capillaries, where they acquire elongated cylindrical shapes. (B) Dual-labeled cancer cells (red: nuclei; green: cytoplasm) trapped inside a blood capillary in a living mouse. The cells were imaged 1 h after injecting them in the epigastric cranialis vein in an abdominal skin flap. The cells and nuclei are deformed by elongation to fit the small (∼8 μm) inner diameter of the capillary. Yellow arrow indicates the outer wall of the blood vessel. Red arrows point out entrapped and elongated human HT-1080 dual-color cells. The right red arrow indicates a cell that has likely undergone mitosis, as judged by its binucleate appearance. Image taken from Suetsugu et al., 2013. 6 Reprinted with permission from ref 6. Copyright 2013 The International Institute of Anticancer Research. (C) 3D schematic of tubular confinement, indicating 2D constraints (red arrows) on 3D cultured mammalian cells. (D) Cell shape factor plotted against microtube sizes (n = 15, 16, 24, 25, 12, and 10 cells from left to right). The inset shows an ellipsoidal fit in orange and black dashed lines for the major (l) and minor (d) ellipsoid axes. NS, not significant; all other cell shape factors are significantly altered compared to free cells (p-value < 0.05; Student's t-test). (E) 3D projected fluorescent microscopy images (histone H2B-mCherry, red; GFP-tubulin, green) of prometa-/metaphase HeLa cells inside differently sized rolled-up nanomembranes (regions between the parallel, dashed lines). (F) Side (upper row) and top (lower row) 3D reconstructed images of chromosome rosettes inside different microtube sizes. The 3D schematics on the left indicate the top view angles (blue arrows) for free (top) and confined (bottom) chromosome rosettes depicted in the lower row on the right. Scale bars, 10 μm.
actomyosin ring that is tightly coupled to the plasma membrane 11 and enables cytokinesis. 16 Mitotic rearrangements of the cortex are accompanied by, and tightly linked to, the remodeling of the interphase microtubule network into a bipolar spindle, which itself possesses an intricate 3D architecture. Dedicated motor-protein activities functioning along microtubules help to provide the vectorial forces required to push/pull the duplicated centrosomes apart, allowing the centrosomes to migrate along various 3D trajectories to opposite cell poles. This process occurs in a tightly controlled temporal manner and generates a mature prometa-/metaphase spindle with extremely welldefined geometric dimensions. 17−19 A ring-like arrangement of chromosomes then forms at the nascent spindle, facilitating bipolar attachment of microtubules to the chromosomes' kinetochores and, thus, enabling faithful chromosome segregation and maintaining genomic stability. 17 Notably, genomic instability can lead to cancer as well as other diseases, 20,21 highlighting the importance of genome-protective mechanisms for human health.
Mitotic progression is highly sensitive to external physical influences, such as spatial stimuli and geometric constraints. 22−24 Indeed, by remodeling the actin cortex lying at the interface between an animal cell and its environment, mitotic cells can translate geometric aspects of their microenvironment into spatial information that determines the fate of their daughter cells. For instance, the pattern of cell adhesion can dictate the orientation of the spindle and thus, the positions of the two arising daughter cells. 22,24 Moreover, preventing mitotic cell rounding by planar compression in one dimension from the top/bottom can cause mitotic delays, multipolar spindles, and defects in chromosome segregation. 14,25 In addition, changes in spindle architecture caused by external forces 26 can impair the stability of the bipolar spindle and the positioning of the cell division plane. 14 Collectively, the above findings illustrate tight connections between the cortical cytoskeleton and spindle microtubules and the highly mechanosensitive nature of cell division. However, prior studies were mainly performed on cells growing on flat surfaces that lack the geometric attributes of the highly curved substrates and tubular confinements cells experience inside the body, such as kidney tubules, mammary ducts, gut villi, vessels, muscle fibers, and bone tissue. Despite the in vivo relevance of these microenvironments, only few in vitro models have been developed to mimic the geometry of such structures, for instance, the ducts and acini of mammary glands. 27 Importantly, recent evidence has emerged indicating that cells under tubular confinement elicit distinct mitotic responses to external spatial stimuli that are different compared to those of spatially confined cells growing on planar substrates. 28 Investigating these responses at a molecular level is therefore crucial to understand better how the physical parameters of different 3D surroundings, such as the tubular confinement of cancer cells inside blood capillaries ( Figure 1B), might dictate mitotic cell behavior and function.
To address this challenge, we have developed and used rolled-up SiO/SiO 2 nanofilms to encapsulate live, individual human cancer cells (HeLa cells) inside transparent microtubes 28−30 ( Figure 1C). This tubular nanomembrane system offers a way of mimicking aspects of the in vivo 3D microenvironments certain cancer cells encounter during metastasis. The behavior of the entrapped cells can then be analyzed conveniently with high-and super-resolution micros-copy. Moreover, due to the on-chip nature of the platform, numerous cells can be imaged in a single experiment. Other systems, such as microfluidic devices, either have complex macro-to-micro interfaces, introduce contamination via oil molecules, have rectangular rather than circular channels, or cannot be implemented in an easily visualizable chip-based format. 31−33 In this study, we show that tubular confinement has a profound impact on a wide range of mitotic features, and we provide an in-depth characterization of these features at a molecular level by using confocal and super-resolution 3D structured illumination microscopy (3D SIM). Thus, we observed deformation of chromosome rosettes into densely packed and disordered shapes and show that tubular confinement influences the kinetics of spindle formation. In addition, consistent with the link between the mitotic spindle and the cell cortex, we discover a striking bipolar redistribution of cortex proteins that occurs specifically in mitotic cells dividing inside nanomembrane tubes. This redistribution coincides with conspicuous membrane blebbing that occurs at the actin-and cortactin-enriched cell tips throughout mitosis. Finally, we establish that membrane blebbing and/or associated processes can act as a potential genome-safety mechanism that becomes functionally important under certain spatial confinements. Based on our findings, we discuss the potential of our platform for uncovering cellular pathways relevant for cells dividing in certain 3D microenvironments. These advances might help provide a better understanding of the molecular events happening, for instance, during early intravascular proliferation of circulating tumor cells.

RESULTS AND DISCUSSION
Tubular Confinement Impacts on Mitotic Cell Rounding, Spindle Morphology, and Chromosome Arrangements. To gain insights into how spatial confinement influences key mitotic features of cells dividing in 3D contexts, we fabricated arrays of on-chip microtubes, inside which individual proliferating human cells, such as HeLa cells, can be entrapped and cultured 28 ( Figure 1C). Briefly, SiO/SiO 2 nanobilayers of 25−100 nm thickness were deposited on a transparent substrate patterned with 100 × 100 μm squareshaped sacrificial layers. After selective etching away of the sacrificial layers, the SiO/SiO 2 nanobilayers were self-folded into microtubes in on-chip format with a density of ∼500 microtubes per 1 cm 2 chip area. 34−36 The diameters of microtubes are highly defined and tunable. 30 We used 7−21 μm microtubes in our initial experiments, because 7 μm represents about one-third of the diameter of rounded-up human HeLa cells (∼20 μm) and because we found that nanomembrane tubes of this size are the smallest tubes that HeLa cells can be entrapped in (not shown). By contrast, tubes with diameters above 21 μm do not impose any spatial confinement on mitotic HeLa cells, as they are larger than the rounded-up diameter of HeLa cells during mitosis. The surfaces of the microtubes were functionalized with fibronectin to mimic aspects of the chemical microenvironment of mammalian cells in vivo. We then co-cultured HeLa cells with the nanomembrane arrays overnight, allowing the cells to spontaneously migrate into the tubular nanomembranes. 28 Finally, we analyzed cells entrapped inside the tubes with high-resolution fluorescence microscopy (confocal microscopy) and/or superresolution 3D SIM. As described below, these techniques

ACS Nano
Article allowed us to gain insights into key processes occurring during single-cell division under tubular confinement.
The rolled-up nanomembrane configuration enables tubular 2D-confinement of 3D cultured mitotic human cells ( Figure  1C). As such, the nanomembrane platform is fundamentally different from previous 1D-confined (from top and bottom) experimental set-ups based on cells cultured on 2D/planar surfaces 14,25 (Supplementary Figure 1). We observed that confinement inside tubular nanomembranes prevented cells from rounding up during mitosis. Instead, cells adopted elongated cylindrical shapes, as illustrated by the strong correlation between the cell shape factor (ratio between the minor and major axes of an ellipse fitted to the cell) and the microtube diameter ( Figure 1D). We therefore focused our subsequent analyses on microtube diameters ranging from 7 to 18 μm, the largest microtubes that induced significant cell shape changes ( Figure 1D). Significantly, this diameter range is comparable to the widths of in vivo blood arterioles/capillaries present in humans and other mammals. 5,6,37 While blood capillaries readily allow passage of endogenous cells of the vascular system, such as red blood cells (∼7 μm in diameter), metastasizing cancer cells that are much larger in diameter (∼20 μm) can get stuck due to size restriction. 4,5 Importantly, cancer cells trapped inside the vascular system adapt to the widths of blood capillaries in vivo, 37 comparable to HeLa cells encapsulated by narrow rolled-up nanomembranes. Thus, the extent of deformation of cancer cells inside blood capillaries in living organisms matches well with the changes in cell shape we observed inside rolled-up nanomembranes: cell shape factors reached ∼0.08 in the very narrowest vascular capillaries inside living mice, 37 comparable to the ∼0.10 cell shape values we determined for HeLa cells inside 7−9 μm rolled-up nanomembranes ( Figure 1D). Collectively, these findings illustrate that the experimental conditions in our system are within the range of tubular constrictions certain cells encounter in vivo.
To probe how tubular confinement affects the mitotic machinery in 3D cultured cells, we investigated how the obstruction of cell rounding influences spindle morphology and chromosome alignment in mitotic prometaphase and metaphase (named prometa-/metaphase henceforth). Both spindle and chromosome alignments are geometrically demanding features that are important for successful mitotic progression. 11,38−40 For our experiments, we used HeLa cells stably coexpressing the core histone H2B tagged with the fluorescent marker mCherry (H2B-mCherry) and α-tubulin fused to green fluorescent protein (GFP-tubulin). This allowed us to simultaneously follow chromosome and spindle dynamics by fluorescence microscopy together with morphological cell shape changes ( Figure 1E). The acquired images revealed that tubular confinement often led to misaligned chromosome showing the temporal evolution of spindle length (mean 3D distances between centrosomes ± SD; time '0' defined as the beginning of cell rounding) in the indicated situations. The ①s with arrows indicate the times at which full separation of the centrosomes is achieved in prometa-/metaphase, i.e., when a mature prometa-/metaphase spindle has formed (black ①: 10.3 ± 0.8 min for free cells and red ①: 32.3 ± 2.1 min for cells in 7−9 μm microtubes). The onset of anaphase spindle elongation is marked by a black and red ② for free cells and cells confined in 7−9 μm microtubes, respectively. In each case, the experiment was repeated at least five times. (C) Quantification of prometa-/metaphase spindle lengths ± SD in the indicated spatial environments (n = 7, 6, 6, 5, and 5 cells from left to right). Blue dotted line indicates the average spindle length corresponding to free cells. (D) Bar graph illustrating average elongation speeds of nascent spindles in indicated spatial environments (blue columns; n = 7, 6, 6, 5, and 5 cells from left to right) and the average time required to form an 11.5 μm long spindle, the average length of prometa-/metaphase spindles in free cells (orange columns; n = 7, 6, 6, 5, and 5 cells from left to right; mean ± SD; see Supplementary Figure 3C for more details). NS, not significant, *significantly altered changes compared to free cells (p-value < 0.05; Student's t-test).

ACS Nano
Article plates and variations in the angles between metaphase plates and spindle axes ( Figure 1E). 3D chromosome reconstructions highlighted a strong compression of prometaphase chromosome arrangements inside microtubes below 13 μm in diameter ( Figure 1F and Supplementary Movies 1−4). During early prometaphase, unconfined HeLa cells dividing on planar surfaces outside of the microtubes (named "free cells" henceforth) displayed typical chromosome rosettes, ringshaped chromosome formations of ∼10 μm in diameter (extracted from our live-and fixed-cell observations). The rosettes consistently formed at the surface of nascent spindles, with the arms pointing outward and the centromeres inward toward the spindle axis, in accord with previous reports 17 ( Figure 1F; Supplementary Figures 2A and B, left and Supplementary Movie 1). These rosettes are believed to be crucial for timely establishment of bipolar kinetochore attachments to opposite spindle poles, a process that relies on a functional prometaphase spindle and is important for faithful sister chromatid segregation. 17,41 Chromosomes captured from this ring by the microtubules are concentrated to form a metaphase plate at an angle perpendicular to the long axis of the spindle. 17 Only when all kinetochores are properly bioriented and the spindle assembly checkpoint (SAC) is satisfied, do cells enter anaphase, triggering a loss of sister chromatid cohesion and the movement of sister chromatids to opposite cell poles. 42 By contrast, 3D reconstructions of cells confined inside microtubes of <14 μm diameters did not contain chromosome rosettes. Instead, their chromosomes exhibited dense packaging into tilted plates lacking a central cavity ( Figure 1F and Supplementary Movies 2−4). These effects were pronounced most strongly in nanomembrane tubes narrower than 10 μm, the mean diameter of chromosome rosettes in free HeLa cells. In those narrow microtubes, the prometaphase chromosome arrangements appeared twisted and strongly elongated into "cloud-like" structures ( Figure 1F Tubular Confinement Impacts on the Dynamics of Centrosome Separation. Chromosome rosettes are important for the timely formation of bipolar prometa-/metaphase spindles. 17 Thus, we wondered whether tubular confinement impacts on the kinetics of spindle formation. To address this question, we took advantage of a Fiji plugin (http://fiji.sc; Fiji is an open source software that is a bundled version of Image J) to automatically monitor the 3D motion of centrosomes during mitosis, using GFP-tagged centrin1 as a marker (centrin1-GFP; Figure 2A, Supplementary Figures 3A and B and Supplementary Movies 5 and 6). The evolution of spindle lengththe distance between the two centrosomesrevealed three distinct phases during mitosis (I−III; Supplementary Figure 3C). This allowed us to determine the timing of the transition from an

ACS Nano
Article elongating nascent spindle to a prometa-/metaphase spindle, which is fairly constant in length ( Figure 2B and Supplementary Figure 3C). The ①s with arrows indicate the times at which full separation of the centrosomes is achieved in prometa-/metaphase i.e., when a mature prometa-/metaphase spindle has formed ( Figure 2B and Supplementary Figure 3C). Moreover, the data indicated the timing for initiation of the final elongation phase of the spindle during anaphase ( Figure  2B and Supplementary Figure 3C, time points indicated with ②). We found that tubular confinement prolonged the formation of prometa-/metaphase spindles from 10.3 ± 0.8 min in free cells (black ①, Figure 2B) around 3-fold to 32.3 ± 2.1 min in cells dividing inside narrow microtubes below 10 μm (red ①, Figure 2B The delays in intermediate and narrow sized tubes were significant compared to free cells (p-value < 0.0009, Student's t test) and could be explained by either or both of two phenomena: first, elongated prometa-/metaphase spindles; and second, slower elongation speeds of nascent spindles. To address the first possibility, we measured the lengths of prometa-/metaphase spindles in cells exposed or not exposed to tubular confinement. Ensuing analyses revealed that, in <10 μm and 10−12 μm microtubes, prometa-/metaphase spindles were ∼4 μm longer than those of free cells (Figures 2B and C, and Supplementary Figures 3B−D). To address the second possibility, we compared the elongation speeds of nascent spindles in cells dividing without or with tubular confinement. Experiments revealed that the mean elongation speed of nascent spindles was significantly reduced from 0.94 ± 0.06 μm/min in free cells to 0.39 ± 0.05 μm/min in 7−9 μm  17 We conclude that the observed spindle formation delays inside rolled-up nanomembranes are likely due to a combined effect of elongated prometa-/metaphasespindle lengths and decreased elongation speeds during spindle formation. Moreover, these elongated prometa-/metaphase spindles inside narrow microtubes plateaued for extensive time periods before further elongating in anaphase ( Figure 2B and Supplementary Figures 3C and D). These findings thus demonstrate that increased durations of spindle formation as well as strongly prolonged prometa-/metaphases contribute to the extensive overall mitotic delays occurring inside rolled-up nanomembranes. 28 Bipolarization of the Mitotic Actin Cortex Inside Rolled-Up Nanomembranes. Mitosis is a geometrically demanding process. As such, it is spatially controlled to ensure that two equivalent daughter cells are produced with high fidelity and in various situations placed into spatially favorable positions. Cells can sense physical cues in their environment to adjust their cytoskeleton. 43 Thus, we tested whether and/or how the cytoskeletal cortex of mitotic cells was affected by tubular confinement. To do so, we stained different mitotic stages of free and confined HeLa cells for actin, a major cytoskeletal protein. Based on confocal and 3D SIM approaches, we performed high-and super-resolution imaging

ACS Nano
Article of actin filaments (F-actin). In free HeLa cells, the overall distribution of F-actin at the cortex (green, labeled by phalloidin) remained uniform in different mitotic stages until the formation of the contractile ring in telophase (see Figure  3A for representative images). By contrast, in confined cells we detected a remarkable enrichment of cortical F-actin at the cell tips, facing the open ends of the microtubes. This bipolar accumulation was reproducibly observed from nuclear envelope breakdown (NEB) onward until telophase, and was specific to mitosis, since no such enrichment was present in confined interphase cells ( Figure 3B, left image). Moreover, the bipolar distribution of actin did not affect formation of the contractile ring in telophase. A similar staining pattern was observed for cortactin, a monomeric protein involved in actin polymerization. Thus, it appears that, under tubular confinement, related, but distinct, cytoskeletal proteins are rearranged in a similar fashion to one another (Supplementary Movie 7). To quantify this redistribution, we measured the fluorescent intensities of phalloidin in single 3D SIM cross sections of mitotic HeLa cells. This revealed a more than 2-fold higher Factin accumulation at the cell tips ( Figure 3C, red arrowhead) compared to the sides underlying the microtube walls ( Figure  3C, light-blue arrowhead). The fluorescence intensities along 0.5 μm xy cortex sections at different locations and their evolution through mitosis are presented in Figure 3D Figure 5).
Conspicuous Membrane Blebbing of Mitotic Cells Inside Rolled-Up Nanomembranes. The bipolar enrichment of actin and cortactin inside rolled-up nanomembranes was accompanied by conspicuous membrane blebbing ( Figure  3B and C, left), with the cell membrane at the tips of the confined dividing cells usually undergoing multiple dynamic cycles of bleb formation and retraction (Supplementary Movie 8). Furthermore, membrane blebbing occurred exclusively during mitosis. Blebs are extensions of the cell membrane caused by contractions of the actomyosin cortex to release hydrostatic pressure. 44 In cells dividing under tubular confinement, pressure can only be released at the cell tips that face the two openings of the microtubes. Interestingly, we observed that blebbing started to take place at much earlier mitotic stages in confined cells than in free cells ( Figure 4A). Based on timelapse images of HeLa cells, we noticed that membrane blebbing in confined situations started in prometa-/metaphase and continued throughout cell division ( Figure 4A, left two images, and Supplementary Movies 8−10). By contrast, in free cells it occurred exclusively during later cell division stages, mainly cytokinesis, and to a much lesser extent ( Figure 4A, right two images, and Supplementary Movie 11). To quantify the spatial positions of membrane protrusions over time, we generated kymographs based on phase-contrast time-lapse images. As shown in the top kymograph in Figure 4B, bleb protrusions (white arrowheads) inside 13 μm microtubes reached lengths of up to ∼10 μm in prometa-/metaphase and more than 35 μm in anaphase (measured from the center of the cell). By contrast, we detected no blebs in free cells in prometa-/metaphase.

ACS Nano
Article Moreover, the farthest membrane protrusions were limited to ∼7 μm in length in anaphase of free cells ( Figure 4B, bottom kymograph). Notably, conspicuous membrane blebbing was also observed in confined 3D cultured mitotic RPE1 cells, 28 demonstrating that the observed phenomena are conserved between different types of human cells. Blebbing at the cell tips (the parts of the cell pointing toward the open ends of the tubes) suggested that cells frequently were under internal hydrostatic pressure inside the nanomembranes.
We found that the membrane blebbing and bipolar actin filament distributions inside microtubes were largely prevented by treating cells with blebbistatin, 45 an inhibitor of the forcegenerating mechanoenzyme myosin II (Figures 4C and 5A, and Supplementary Movies 12−14). Membrane blebs have recently been implicated in mitotic functions, being shown, for example, to contribute to the control of spindle positioning 46 and to act as a partly redundant genome safety mechanism by stabilizing the cleavage furrow during cytokinesis. 47 Indeed, we found that in confined cells, inhibition of blebbing and/or associated processes caused misplacement of the cleavage furrow during cytokinesis and often led to unequally sized asymmetric daughter cells (Supplementary Figure 6A Figure 7). This discrepancy could reflect differences in cell lines between the two studies (HeLa cells here versus PtK2/B6−8 cells in the previous report) 48 or the 4fold lower blebbistatin concentrations we used (25 μM versus 100 μM in the previous report). 48 Thus, the lower concentration of blebbistatin in our study may inhibit some (e.g., membrane blebbing) but not all myosin II-mediated processes. Collectively, our data suggest that blebbing and/or associated processes contribute to accurate chromosome segregation, particularly in cells under tubular confinement. Under these conditions, blebbing and/or associated processes act as a compensatory mechanism to limit the deleterious effects of spatial confinement on chromosome segregation and genome stability.

CONCLUSIONS
We have used "rolled-up" nanotechnology to generate transparent biocompatible 3D structures of tailored tubular geometry with varying diameters. These nanomembrane tubes can be integrated into an experimentally convenient pipeline to investigate the growth and proliferation of mammalian cells under spatial constraints at high resolution. The SiO/SiO 2 bilayer nanomembranes are therefore viable substrates for studying the effects of tubular confinement on various cellular processes including but not limited to geometrically demanding processes such as cell division. The "rolled-up technique" that we have employed integrates nanofilm engineering and standard photolithography processes 34 and can be readily scaled-up to produce microtubes as on-chip arrays. The dimensions (lengths and diameters) and wall thicknesses of these architectures can be precisely controlled, and moreover, the inner surfaces can be functionalized with biomolecules, such as fibronectin. The ease of combining our platform with highand super-resolution microscopy allowed us to reveal and analyze mitotic alterations under narrow tubular confinement, namely: (1) abnormal spindle morphology including increased lengths of prometa-/metaphase spindles; (2) distorted chromosome arrangements in prometaphase that are in stark contrast to the highly ordered chromosome rosettes occurring in unconfined situations; (3) delays in centrosome separation dynamics during spindle formation; (4) bipolar distribution of the actin/cortactin cell cortex; and (5) enhanced mitotic membrane blebbing that occurs at the cell tips facing the open ends of the microtubes.
Numerous biological systems achieve robustness by employing feedback loops. 49,50 Our observations suggest that the mechano-and/or geometry-sensing system of redistributing proteins in response to spatial constraints may be part of such a feedback mechanism that monitors cell shape and/or associated features during mitosis. Indeed, we speculate that cells may sense and respond to spatial strain in the cortex, perhaps by opening ion channels or by stretching cortical cytoskeletal proteins such as actin to create new binding sites 51,52 and triggering the recruitment of contractile/cortical and/or other proteins to further translate the signal. Such scenarios will be interesting to investigate in future studies.
The effects of tubular constraints that we have observed on key mitotic features such as the spindle, chromosome rosettes and centrosome separation, are consistent with the previously reported impact of spatial constraints on mammalian cell division. 14,25 Given that perturbations of spindle morphology and dynamics impair the "search-and-capture" interactions between microtubules and kinetochores, 53,54 our findings could help explain the subsequent prometa-/metaphase delays and the increased chromosomal instability observed previously in such contexts. 28 Indeed, the information presented herein provides valuable clues as to what potential causes could trigger the high percentage of chromosome segregation errors (CSEs) occurring inside microtubes. Intriguingly, our findings suggest that some mitosis-specific responses, especially membrane blebs, are involved in limiting chromosome segregation errors in certain confined environments. Such responses might also

ACS Nano
Article operate within the context of certain steps during metastasis when large cancer cells undergo tubular confinement inside blood arterioles and capillaries. It will thus be interesting to explore precisely how mitotic processes as diverse as SAC activation, cortex bipolarization, and membrane blebbing are orchestrated to help limit genome instability specifically in cells dividing in 3D environments under such types of spatial confinement. Our 3D nanomembrane technology could provide an excellent platform for such future studies. Notably, evidence has emerged that primary and secondary tumors can vary in their genomic/chromosomal makeup. 8−10 It will be interesting in the future to see whether aberrant mitotic processes, occurring during intravascular proliferation, could in part explain these differences.
In summary, our study provides an in-depth characterization of a versatile rolled-up nanomembrane platform that can be used as a powerful tool for live single-cell studies. Indeed, the system can be applied to virtually every cellular pathway amenable to high-and/or super-resolution microscopy. Rolledup nanomembranes are highly adaptable in their design and could easily be combined with various other technologies, such as electronic circuits and microfluidic lab-in-a-tube systems. 29 Moreover, we believe that the array setup of the tubular nanomembranes has considerable potential to be exploited for tissue engineering/development and high-throughput drug screening.

METHODS AND EXPERIMENTAL SECTION
Fabricating Microtubes in On-Chip Format for Mammalian Cell Culture. We fabricated microtubes by rolling-up nanomembranes on square glass slides (18 mm in length; thickness, 170 μm). To prepare sacrificial layer patterns, a 2.4 μm thick positive photoresist (AR P-3510, Allresist GmbH) film was first prepared on the glass slides by spin-coating at 3500 rpm for 35 s followed by a baking step at 90°C for 5 min on a hot plate. A conventional photolithography step using a contact mode mask aligner (MA56, SUSS MicroTec AG) was exploited to produce designed patterns on the glass slides. After that, SiO and SiO 2 nanomembranes were deposited onto the patterns with an electron beam evaporation system (BOC Edwards FL400, Germany) at a glancing angle of 30°and deposition rates of 5.0 Å/s for SiO and 0.5 Å/s for SiO 2 , respectively. To achieve microtubes with diameters ranging from 4 to 20 μm, SiO/SiO 2 bilayers of thicknesses between 25 and 100 nm were deposited. The binanomembranes were immediately rolled-up into microtubes by under-etching the sacrificial layers in an acetone solution for a few seconds. The samples were dried by a super critical point dryer to prevent the tubes from collapsing. To strengthen the microtubes, they were further coated with an Al 2 O 3 layer of 15 nm in thickness, using atomic layer deposition (Savannah 100, Cambridge NanoTech Inc.) at 80°C.
Functionalization of the Microtubes with Fibronectin. To grow a monolayer of octadecanylphosphonic acid on the microtube arrays, the samples were incubated in a toluene solution (50 μM) overnight, followed by a thorough rinse with toluene, acetone and water. The samples were then immersed into a 1× PBS solution, containing 100 mM N-(3-(dimethylamino)propyl)-N′-ethylcarbodiimide hydrochloride (EDC) and 25 mM N-hydroxylsulfosuccinimide (NHS) for 1 h followed by an incubation step with 20 μg mL −1 fibronectin for 1 h at 37°C.
Inhibitor Treatments. Blebbistatin (Sigma-Aldrich) was used at a final concentration of 25 μM and was added 1 h before cells entered mitosis.
Live-Cell Imaging. For time-lapse imaging, the microtube arrays were mounted on 35 mm dishes (MatTek Corporation). The cells were then plated onto these arrays and treated with/without chemical reagents according to the method mentioned above. Imaging was performed in a heated and humidified chamber (stabilized at 37°C and containing 5% CO 2 ). Cells were imaged, using an Axio Observer Z1 inverted microscope (Zeiss, 40× objective). For time-lapse imaging of cell division events, we captured one frame every 5−15 min, depending on the specific experiment. Z-stacks (0.5−1.0 μm) of the mitotic apparatus (entire volume) were obtained every 5 min with a PLANAPO 40× objective. The images were deconvolved using Huygens and analyzed using Fiji. 58 Images represent maximumintensity projections of all z-planes.
High-and Super-Resolution Imaging. Entrapped and free HeLa cells were fixed in prometaphase and, in the case of untagged HeLa cells, stained with DAPI and phalloidin−FITC. Confocal imaging was performed with an LSM 700 inverted microscope (Zeiss) equipped with a Zeiss C-Apochromat 60 × 1.2 W objective, using 405, 488, and 594 nm laser lines for excitation, a 1532 × 1532 pixel frame size at a rate of 0.42 μs per pixel and a 1 Airy Unit pinhole diameter. Z-sections taken every 410 nm covering the entire cell were projected (maximum intensity).
Super-resolution 3D SIM imaging of phalloidin-FITC (F-actin) and DAPI as well as H2B-mCherry and GFP-tubulin was acquired using a Deltavision OMX 3D SIM System V3 from Applied Precision (a GE Healthcare company) equipped with 3 EMCCD Cascade cameras from Photometrics, 405, 488, 592.5 nm diode laser illumination, an Olympus PlanSApo 100× 1.40 NA oil objective, and standard excitation and emission filter sets. Imaging of each channel was performed sequentially using three angles and five phase shifts of the illumination pattern as described in Gustafsson et al., 2008. 59 The refractive index of the immersion oil (Cargille) was adjusted to 1.514 to minimize spherical aberrations. Sections were acquired at 0.125 μm z steps.
Raw OMX data were reconstructed and channel registered in the SoftWoRx software (Applied Precision, a GE Healthcare company). Reconstructions were carried out using channel-specific Optical Transfer Functions (OTFs) and channel-specific K0 angles. OTFs were generated within the SoftWoRx software by imaging 100 nm beads (Life Technologies) using appropriate immersion oils to match the data. Channel registration was carried out using the Image Registration parameters generated within the SoftWoRx software and checked for accuracy by imaging Tetraspeck beads (Life Technologies). Channel registration was accurate to within one pixel. Further data analysis was performed using Fiji. 58

ACS Nano
Article 3D Reconstruction of Chromosomes. Z-stack images were deconvolved with Huygens Professional 4.1 (SVI), and 3D reconstruction was performed by its module surface renderer.
Statistics and Computational Analysis. Means ± standard deviations presented herein were obtained by comparing at least three independently performed experiments. Student's t-tests, as indicated in the figure legends, were performed to evaluate statistical differences between two groups, as indicated in the Figures. P-values ≥ 0.05 were considered nonsignificant.
The 3D Euclidean distances between centrosomes were calculated using Ace 3D, a visualization and analysis plugin for Fiji 58 that uses 3D Object Counter 60 to create multidimensional centroid maps for analysis (available on request).
The 3D temporal evolution of spindle lengths (mean distances between centrosomes ± SD) was plotted against time using