Microtubule-dependent pushing forces contribute to long-distance aster movement and centration in Xenopus laevis egg extracts.

During interphase of the eukaryotic cell cycle, the microtubule (MT) cytoskeleton serves as both a supportive scaffold for organelles and an arborized system of tracks for intracellular transport. At the onset of mitosis, the position of the astral MT network, specifically its center, determines the eventual location of the spindle apparatus and ultimately the cytokinetic furrow. Positioning of the MT aster often results in its movement to the center of a cell, even in large blastomeres hundreds of microns in diameter. This translocation requires positioning forces, yet how these forces are generated and then integrated within cells of various sizes and geometries remains an open question. Here we describe a method that combines microfluidics, hydrogels, and Xenopus laevis egg extract to investigate the mechanics of aster movement and centration. We determined that asters were able to find the center of artificial channels and annular cylinders, even when cytoplasmic dynein-dependent pulling mechanisms were inhibited. Characterization of aster movement away from V-shaped hydrogel barriers provided additional evidence for a MT-based pushing mechanism. Importantly, the distance over which this mechanism seemed to operate was longer than that predicted by radial aster growth models, agreeing with recent models of a more complex MT network architecture within the aster.

INTRODUCTION

The microtubule (MT) cytoskeleton is critical for establishing intracellular organization. During interphase, this network adopts the form of an astral array with MTs emanating from a central nucleating focus (centrosomes in animal cells) and branching outward throughout the cell. Importantly, the center of this aster is typically found near the geometric center of the cell, where it associates with the surface of the nuclear envelope. Forces generated by the MTs themselves move organelles during interphase, and because the network serves as a scaffold upon which these organelles attach and move, the aster’s position dictates their spatial arrangement within the cell. During mitosis, spindle assembly is initiated in part by MT nucleation from centrosomes, and thus centrosome location at the interphase-to-mitosis transition has a role in determining where the spindle forms and in establishing the eventual location of the cell division plane (for a review, see McNally, 2013).

Although aster centering is likely to be critical in all cells, it is particularly relevant in large single-celled blastomeres immediately after fertilization. Here, the male pronucleus and its associated centrosome must traverse large distances to reach the cell center and establish the location of pronuclear fusion, spindle formation, and division plane positioning during the subsequent mitosis (e.g., Reinsch and Gonczy, 1998; Wuhr ). Thus, errors in this process can lead to erroneous cell division and have deleterious effects on developmental progression. Precisely how the MT aster generates and responds to forces to move to the cell center remains unanswered, in part because the most commonly used biological model systems—large blastomeres from sea urchins (Minc ; Tanimoto ), Caenorhabditis elegans embryos (Gonczy ; Labbe ; Nguyen-Ngoc ), sand dollars (Rappaport and Rappaport, 1994), and Xenopus laevi (Wuhr , 2010)—are intrinsically limited in terms of the shapes they can adopt (although some impressive work to manipulate shape has been done in these systems, e.g., Minc ; Chang ; Tanimoto ). With the notable exception of C. elegans embryos, the aforementioned model systems are also limited in terms of their genetic and biochemical tractability. Furthermore, although large cells are somewhat amenable to imaging, their thickness precludes live-cell imaging with a high signal-to-noise ratio that would allow for a more detailed characterization of the spatial and temporal dynamics of the MTs involved. Addressing these limitations will be important for advancing our current understanding of the MT cytoskeleton, including how it moves within the cell and how it organizes the cytoplasm.

Most studies addressing aster positioning have typically approached the phenomenon in one of two ways, either by using purely reconstituted systems confined to small wells in polydimethylsiloxane (PDMS) microfabricated chambers (Holy ; Dogterom ; Roth ) or by manipulating living blastomeres physically and/or biochemically (Kimura and Kimura, 2011; Minc ; Tanimoto ). These studies are complemented by theoretical descriptions and models of the process (e.g., Letort ; Nazockdast ). However, there still exists a need to differentiate between possible underlying biophysical mechanisms. Currently, there are three prevailing models used to describe how an aster can generate (or respond to) the forces necessary to move it from the edge of a large cell to the center of that cell, all of which are equally plausible and not mutually exclusive: cortical pulling, cortical pushing, and cytoplasmic pulling (Nazockdast ; Wu ).

In the cortical pulling model, minus end−directed motors that are bound to the cell cortex (e.g., cytoplasmic dynein) generate pulling forces on the plus ends of MTs as they interact with the cell cortex (Laan , b; Lammers and Markus, 2015). However, this model fails to adequately explain other experimental observations, including results from MT ablation/severing studies (Tanimoto ), ;which indicate that the nucleating center of the aster moves to follow the center of its MT mass (CoMM), a result more consistent with the cytoplasmic pulling model (see below). Additionally, in the context of aster positioning in large blastomeres, aster movement away from the proximal cortex begins before MTs reach the opposite side of the cell (Wuhr , 2010), suggesting that other mechanisms are likely responsible, at least at the initial stages of aster movement away from a barrier.

Like the cortical pulling model, the cortical pushing model involves interactions between MTs and the cell cortex. In this case, however, the driving force is generated by polymerizing MT ends as they grow into the rigid “barrier” represented by the cell cortex. This model is supported by in vitro observations and measurements of forces generated by MT−barrier interactions (Holy ; Laan ) as well as by in vivo observations of MT buckling (Brangwynne ) and in small cells such as yeast and Dictyostelium (Tran ; Brito ). The validity of this model is undermined by the fact that the mechanical properties of individual MTs make them poor bearers of compressive loads and that the critical length of MT buckling is an order of magnitude or two shorter than the distances that asters traverse in large cells (Brangwynne ). However, the architecture and connectivity of the MT network within an aster, in terms of MT bundling and branching, could dramatically affect the length scale over which an aster could exert productive pushing forces against a barrier.

Finally, there is the cytoplasmic pulling model, which describes forces thought to be generated by the movement of cytoplasmic dynein (hereafter “dynein”), a minus end−directed MT motor, as it transports cargo from the peripheral cytoplasm toward the aster center (Kimura and Onami, 2005; Tanimoto ). As the dynein motors and associated cargoes move through the viscous cytoplasm, the drag force generated is countered by an equal and opposite force on the aster. When integrated over the entirety of the aster, these forces are sufficient to move the aster. This model predicts that force should be proportional to asymmetries in MT density (a proxy for motor density) and is supported by recent experimental observations of dynein-dependent aster motion in large blastomeres (Kimura and Kimura, 2011; Tanimoto ) and by compelling observations of aster movement in sea urchin embryos following induced aster asymmetry–during which the asters always moved away from the site of laser-induced MT ablation/severing (Tanimoto ).

Here we describe the application of a new approach engineered to overcome existing limitations inherent to earlier methods and to investigate the biophysics of aster movement and centration in large cells. It combines PDMS microfluidic devices, photolabile hydrogels, X. laevis egg extract (Desai ; Maresca and Heald, 2006; Good and Heald, 2018), and confocal microscopy to resolve the mechanics of aster movement in a tractable, cell-free system that closely resembles the cytoplasm within a cell (Parsons and Salmon, 1997; Desai ). Using this approach, our observations of aster movement and centration in confining geometries suggest that MT-dependent pushing forces are sufficient to move asters over the length scales explored and that the arrangement of MTs within the aster is capable of bearing the resulting compressive loads.

RESULTS

Aster centration in PDMS channels does not require dynein activity

To investigate the mechanics of aster centration in a way that allows greater flexibility in exploring the effects of enclosure geometry, we confined X. laevis egg extract containing artificial microtubule-organizing centers (aMTOCs) in either PDMS walled channels (Figure 1, A and B; Supplemental Videos 1 and 2, respectively) or annular cylinder hydrogel enclosures within larger PDMS channels (Figure 1C; Supplemental Video 3) and monitored their position over time using time-lapse spinning-disk microscopy. The poly(ethylene glycol)-based hydrogels used in these experiments have proven to be compatible with X. laevis egg extracts (Supplemental Figures S1 and S2; Bisht ; Geisterfer ) and exhibit low levels of nonspecific protein binding to their surfaces (Nuttelman ; LeValley ). We also note that our extracts contained cytochalasin D (10 µg/ml (Desai ), which is routinely added during preparation to disrupt the formation of filamentous actin and inhibit gelation and contraction of the extract (Field ). As such, the physical properties of the extracts used here likely differ from those found in vivo. MTs were visualized by supplementing egg extract with an engineered fluorescent MT-binding protein (mCherry-TMBD; Mooney ). This visualization strategy was chosen because these tau-based constructs confer a higher signal-to-noise ratio compared with traditional MT-labeling approaches using fluorescent tubulin and have been shown to only minimally affect MT dynamics (Mooney ; Field ; see also Supplemental Figures S1 and S2).

FIGURE 1:

Dynein inhibition via addition of p150-CC1 does not affect aster centration in X. laevis egg extracts. Centration of aMTOCs in interphase X. laevis egg extracts was investigated in PDMS microfluidic channels or hydrogel annular cylinders. Time-lapse spinning-disk confocal microscopy was used to visualize MTs, which were labeled with mCherry-TMBD and are shown in grayscale (A–C). The dashed line in each series of images runs through the midpoint of the channel width (A, B) or the center of the annular cylinder (C) and represents 50% of the channel width or interior diameter accordingly. Images in A show an aster starting near a PDMS channel wall (0%) and centering over time in untreated extract, whereas the images in B show similar aster centration in extract treated with 2 µM p150-CC1. Images in C show aster centration within a hydrogel annular cylinder in untreated extract. For each experimental condition, the aMTOC position (red asterisk) relative to the proximal wall was plotted over time as a percent of the channel width in D (red lines) and E (blue lines) and of the interior diameter in F (green lines). The colored lines represent traces from each individual experiment; n ≥ 10 for each condition. Yellow arrowheads point to MT bending. For all images, scale bar = 20 µm.

Movie S1

Aster centering in a symmetric channel filled with untreated extract.

Movie S2

Aster centering in a symmetric channel with p150-CC1 treated extract.

Movie S3

Aster centering in an annular cylinder filled with untreated extract.

To determine whether we could recapitulate in vivo aster centration as it occurs in large blastomeres (∼100 µm in diameter; Minc ; Tanimoto ), we first flowed extract containing aMTOCs into PDMS devices containing arrays of channels that were ∼1 mm in length, ∼100 ± 10 µm in width, and ∼30 µm in height. Filled devices were kept on ice until being placed on the microscope stage to prevent MT polymerization and movement of aMTOCs. To normalize for the slight variation in channel width among devices, aster movement was tracked as a percentage of channel width, with 0% representing the nearest wall and 50% representing the center of the channel. Only aMTOCs that were within 30% of the nearest wall at experimental onset were tracked. We found that asters in channels were able to center, on average, within ∼20 min and, once centered, tended to maintain their position within the channel (Figure 1, A and D). We also observed MTs fanning out along the proximal wall, which we attribute to some MTs slipping along the surface as they grow. aMTOC asters confined within a more cell-like geometry of annular cylinders made of poly(ethylene glycol) diacrylate (PEGDA) hydrogels with diameters of ∼100 µm also centered and did so with similar velocities as compared with asters in PDMS channels (Figure 1, C and F; Supplemental Figure S3). These data suggest that the underlying mechanism was largely unaffected by the specific composition of the barrier surfaces and that the underlying forces were capable of centering asters over the length scales tested, ∼50 µm, or half the width of our channels and half the diameter of our annular enclosures.

To begin to elucidate the nature of the forces responsible for centration and to differentiate more broadly between a MT-dependent pushing mechanism and one that involves dynein-dependent pulling, we compared the dynamics of aster centration in PDMS channels in the presence and absence of the dynein inhibitor p150-CC1 (2 µM; Gaetz and Kapoor, 2004). When added to spindle assembly reactions in mitotic extracts, this treatment resulted in the well-established “haystack” spindle phenotype (Mitchison ; Gatlin ). This phenotype was also observed in assembly reactions within PDMS channels containing hydrogel structures and in the presence of mCherry-TMBD to visualize MTs (Supplemental Figure S1), suggesting that the reagents used in our studies, as well as the binding of mCherry-TMBD to MTs, likely did not interfere with p150-CC1−mediated dynein inhibition. The same treatment also resulted in a failure of membrane accumulation near the aster center (Supplemental Figure S1B), as observed previously (Hara and Merten, 2015; Cheng and Ferrell, 2019). Taken as a whole, these data suggest that our experimental approach had little effect on the ability of p150-CC1 to inhibit dynein. Interestingly, and in contrast to previous reports in which aster centration was negatively affected after dynein inhibition using ciliobrevin (Tanimoto , 2018; Cheng and Ferrell, 2019), dynein inhibition using p150-CC1 had no noticeable effect on aster centration in our studies (Figure 1, B and E). These data suggest that the centering mechanism used under our experimental conditions does not require the activity of dynein. We note, however, that previous studies have shown that aster centration in X. laevis embryos is indeed sensitive to p150-CC1 treatment (Wuhr ), an apparent conflict that we address in the Discussion.

Aster centration can be disrupted using asymmetric hydrogel structures

We next investigated whether aster centration in extracts can be affected by aster asymmetries induced by the geometry of the confining enclosures. Both the cytoplasmic pulling and the cortical pushing models rely on asymmetries in the aster to induce motion. For the cytoplasmic pulling model, it is asymmetry in the aster itself that generates the force in the direction of the geometric center. In contrast, for a pushing-based model it is an asymmetry in position that causes force to be directed away from a proximal barrier. On the basis of observations of aster shape in our smooth-walled PDMS channels, specifically the fanning-out of MTs along the proximal wall (e.g., Figure 1A, 10-min image), we posited that MT ends might impact the wall and then slide along it. Indeed, we observed slipping of EB1-GFP−labeled MT ends as they grew into hydrogel barriers (Supplemental Figure S4). In these experiments, the tendency of growing MT ends to slip and grow along the barrier seemed to increase as the contact angle with the barrier became more acute. We reasoned that slipping ends, if unable to gain purchase, would produce less pushing force against the barrier. Therefore, if we could mitigate this slipping, we could affect aster centering. To test this hypothesis, we generated channels in which one wall was smooth and the other was sawtoothed, with peaks and troughs designed to reduce MT slippage on one side of the channel (see Materials and Methods; Figure 2A; Supplemental Video 4) We defined the midpoint of these asymmetric channels as the point located halfway between a line running parallel to the channel wall and through the geometric center of the sawtooth peaks (and troughs) and the opposite PDMS wall. This yielded devices with an average width of 100 ± 20 µm. As in Figure 1, channel widths were normalized to account for variation, and aster movement was tracked as a percentage, with 50% corresponding to the channel midpoint (Figure 2A). Only aMTOCs whose starting positions were closest to the sawtooth side of the channel were tracked. Several asters were able to move beyond the channel midpoint and achieve a steady state position farther away from the sawtooth side of the channel and closer to the smooth side, resulting in a noticeable increase in the average deviation from the channel center as compared with the symmetric controls (Figure 2, A and C; Supplemental Figure S3). This observation suggests that a net force, directed away from the sawtooth wall, is applied to the aMTOC at the channel midpoint and is consistent with the idea that growing MT ends can more easily gain purchase on the sawtooth wall.

FIGURE 2:

Asters are unable to center when confined in certain asymmetric enclosure geometries. The ability of aMTOC asters to center was investigated in asymmetric microfluidic channels (A) and asymmetric hydrogel structures designed to resemble the cross-section of the inside of a lobster trap (B). Aster movement was visualized and tracked as described in Figure 1. For each set of time-lapse image series, the dashed yellow lines run through the midpoint of the asymmetric channel width (A) or the center of the lobster trap enclosure (B) and represent 50% of the channel width or interior length, respectively. For the experimental conditions shown in A and B, the aMTOC position (red asterisk) relative to the nearest wall at experimental onset was plotted over time as a percent of the channel width and interior length with the graphs in C (red lines) and D (blue lines), respectively. The colored lines represent individual traces for each experiment; n ≥ 9 for each condition. For all images, scale bars = 20 µm.

Movie S4

Aster moving off center an asymmetric sawtooth channel with untreated extract.

One way to define aster asymmetry is to calculate the distance between a point representing the nucleating center of the aster, in our case the aMTOC, and another representing the center of mass of all aster MTs (referred to as the CoMM). For cytoplasmic pulling models, the larger the separation between these two points, the larger the degree of aster asymmetry and, in theory, the larger the net centering force. In the absence of other types of forces, the net centering force produced by pulling against the cytoplasm should approach zero as these two points converge. As such, if a confined aster could achieve a steady state position in which its nucleating center (or aMTOC) and its CoMM are spatially separated, it would suggest that other types of forces are at play, for example, a countering pushing force. To test this hypothesis, we challenged aster centering within enclosures of a more complex geometry, one that resembles the cross-sectional area of a traditional lobster trap (Figure 2B; Supplemental Video 5). By design, the geometric center of these enclosures lies outside of the bottom chamber of the trap (as oriented in Figure 2), near the 50% midpoint of the trap’s long axis (∼130 µm in length). aMTOCs were initially positioned in the bottom chamber (or in the trap) and their position tracked over time as they nucleated MTs, formed asters, and moved. The position of each aMTOC was graphed as a percent of the length of travel along this long axis (Figure 2D). We found that asters were, without exception, unable to exit the trap and reach the geometric center of the enclosure. Instead, they seemed to reach an equilibrium position near the geometric center of the bottom chamber. This is despite aster MTs extending throughout the entirety of the enclosure, making the position of the geometric center of the enclosure and that of the aster CoMM the same. We argue that the inability of aMTOCs to migrate to the geometric center of these enclosures (and, by extension, to the CoMM of the aster) suggests that a cytoplasmic pulling mechanism is not solely responsible for aster movement and centering in this system.

Movie S5

Aster failing to center in a lobster trap with untreated extract.

Aster movement away from V-shaped hydrogel barriers supports a pushing-based mechanism

To further elucidate the force mechanism(s) responsible for generating aster movement and centration, we investigated the impacts of aster proximity to a wall and of induced aster asymmetry on aster movement. To address this experimentally, aMTOCs were transiently captured in small degradable hydrogel cylinders composed of poly(ethylene glycol)di-photodegradable acrylate (PEGdiPDA), a photolabile polymer derivative of PEGDA (Kloxin ). The unpolymerized PEGdiPDA was then washed out of the microfluidic device and replaced with PEGDA to generate photostable hydrogel barriers near the PEGdiDPA-encapsulated aMTOCs. The end result was an encapsulated aMTOC positioned at the vertex of a V-shaped PEGDA hydrogel barrier with either a 30° or a 90° interior angle (Figure 3A). After unpolymerized PEDGA solution was washed out of the device and replaced with interphase egg extract, the PEGdiPDA cylinder encasing the aMTOC was degraded via exposure to 365-nm UV light, thereby releasing the aMTOC and allowing it to begin nucleating MTs. At this point in the experiment, we began tracking aMTOC motion away from the vertex as a function of time. In contrast to the previous experiments (shown in Figure 1) in which aMTOCs were initially free and positioned somewhat randomly at experimental onset, this approach allowed us to synchronize the starting position and afforded exquisite control over the initial position of the aster as well as the timing of movement onset.

FIGURE 3:

Aster movement away from V-shaped barriers implies a pushing-based mechanism for translocation. Each of the paired sets of cartoon depictions in the top row of A shows a microfluidic device (left) and a zoomed-in view of the channel interior (right) for the different steps required for aMTOC capture and V-shaped structure formation. Corresponding bright-field and fluorescent images of each sequential step, moving left to right, are shown in the bottom row. (i) A photolabile prepolymer hydrogel solution (PEGdiPDA) containing aMTOCs was flowed into the channel of a microfluidic device. (ii) The hydrogel solution was exposed to UV light (λ = 405 nm) patterned using a digital micromirror array placed in a conjugate plane to the specimen in the light path of the microscope. For these experiments, a small circle in the electronic mask was aligned with an aMTOC in the PEGdiPDA solution. Upon exposure, this produced a cylindrical column of photolabile hydrogel surrounding an aMTOC at its base, temporarily fixing it in place and preventing the aMTOC from being washed away in subsequent steps. This anchoring step was repeated for multiple aMTOCs in the prepolymer solution. (iii) A nondegradable prepolymer hydrogel solution (PEGDA) was flowed into and exposed to UV light (λ = 405 nm) in the shape of a V aligned such that the aMTOC being targeted was positioned close to the vertex of the hydrogel structure. (iv) After the generation of all structures, the unpolymerized solution within the device was washed out with buffer and ultimately replaced with interphase X. laevis egg extract. (v) The aMTOC was released by degrading the PEG-diPDA cylinder with exposure to higher-energy UV light (λ = 365 nm). Extracts were supplemented with mCherry-TMBD, and aMTOC aster movement was recorded using time-lapse spinning-disk confocal microscopy as in Figure 1. An example of aster movement from the vertex of a 30°V is shown in B and from a 90°V in C. The yellow dashed line in each image extends from the center of the aMTOC position (red asterisk) in B and is included to facilitate comparison of aMTOC velocity away from the vertex in C. aMTOC movement away from the vertex of the 30°V (average position, red line) and 90°V (average position, blue line) structures was plotted as a function of time to make the graphs shown in D. Shaded outlines represent 95% confidence intervals; n ≥ 10 for each V type. Using the slopes from these plots, the aMTOC instantaneous velocity was plotted vs. the shortest distance to the nearest barrier surface (E) and the distance between the CoMM (see Materials and Methods) and the aMTOC (F) for both 30°V (filled red triangles and squares, respectively) and 90°V (open blue triangles and squares, respectively) structures. Here a positive velocity indicates that the aMTOC was moving away from the vertex. For all images, scale bars = 20 µm.

V-shaped barriers with interior angles of 30° (30°V) and 90° (90°V) were used in these experiments. The specific barrier shape was chosen because it forced the expanding MT network to adopt an asymmetric morphology, being constricted on the side closest to the vertex and more expansive on the open side of the V. These hydrogel structures were made near the middle of large channels (∼1 cm long, 1 mm wide, and 30 µm high) within our PDMS devices, typically hundreds of microns away from each other or any vertical PDMS wall. By varying the angle of the V, we ensured that the aMTOC would always be closer to a barrier in the more acute 30°V as compared with the more obtuse 90°V (see Figure 3, B and C, respectively). For both geometries, released aMTOCs nucleated MTs and began to move away from the vertex along a horizontal line centered on the V with very little deviation from the center of the V (Figure 3, B and C; Supplemental Videos 6 and 7). The distance traveled along this line was plotted as a function of time for each geometry (Figure 3D). On average, the aMTOCs in the 30°V moved faster and farther from the vertex than those in the 90°V.

Movie S6

Aster moving away from the vertex of a 30{degree sign}V with untreated extract.

Movie S7

Aster moving away from the vertex of a 90{degree sign}V with untreated extract.

If modeled as a beam, as MTs push against a barrier the force they can exert before buckling is proportional to 1/(MT length)2. Because the aMTOC is always closer to a barrier wall in the 30°V than in the 90°V at the same distance away from the vertex, if pushing is the responsible mechanism, the force exerted on the aMTOC in the 30°V will be larger than that experienced in the 90°V, perhaps explaining the difference in observed aMTOC velocities (Figure 3D). Indeed, MTs are capable of polymerizing against a hydrogel barrier in our system, based on observations of the interactions between growing MT ends and barriers (Supplemental Figure S4). In contrast to MT pushing models, cytoplasmic pulling models predict that aMTOC velocity should be proportional to the distance between the aMTOC and the aster’s CoMM (see Materials and Methods) and should be directed along the line connecting the two. To attempt to differentiate between these two models, “instantaneous” aMTOC velocities were calculated by taking the slopes of the distance-versus-time plots in Figure 3D at 5-min intervals and plotting them against either 1) proximity to the nearest wall, as determined geometrically (Figure 3E) or 2) the distance between the position of the aMTOC and the aster’s CoMM (Figure 3F). In these plots, positive velocity values were assigned to aMTOC movement away from the vertex (left to right as depicted in Figure 3). In both the 30°V and 90°V barriers, velocities were higher closer to a barrier and dropped off as the aster moved farther away from the vertex, consistent with a pushing model. We found that this relationship held constant for all experimental conditions tested (Supplemental Figure S3). Interestingly, we found that the highest aMTOC velocities correlated with negative distances between the aMTOC and aster CoMM, meaning that aMTOCs moved away from the vertex faster when more MTs were present between the aMTOC and the vertex than were in front of the aMTOC. Particularly at later time points, these geometries produced the expected aster asymmetry, reflected by spatial separations between the positions of the aMTOC and the aster CoMM (Figure 3F). When these spatial separations were plotted versus the instantaneous aster velocity (Figure 3F), we found a negative correlation between the two parameters. This observation suggests that a pulling-based mechanism cannot be solely responsible for aster movement away from a barrier.

Decoupling aster growth from aster movement suggests that bulk translocation of the entire aMTOC-associated MT network is not required for aster centration

Current models of aster centration (both pushing and pulling based) implicitly require bulk translocation of the entire aster MT network as it moves through cytoplasmic space en route to the cell center. Whether this assumption is truly valid remains unaddressed, in part because asters in vivo are typically expanding while they move and achieve centration only coincident with expansion throughout the entirety of the confining cell. The question remains as to whether an aster’s nucleating center can find the center of an enclosed space without bulk translocation of the entire aster MT network, and indeed, recent experimental evidence suggests that this might be the case. For example, a recent in vitro study suggested that asters can simply assemble de novo in confined volumes and effectively center themselves without translocation (Juniper ). Additionally, nuclei, and presumably their surface-associated centrosomes, seem to be able to move independently of asters that form in the tessellated compartments of cycling extracts (Cheng and Ferrell, 2019). To determine whether aMTOCs can find the center of enclosures when aster movement is decoupled from aster growth, we used a modified version of the method outlined in Figure 3A, in which the aMTOC was only partially embedded in a temporary PEGdiPDA cylinder (Figure 4A). This allowed the aMTOC to nucleate MTs from its extract-exposed surface while preventing its movement until it was released by photo-induced degradation of the cylinder. Additionally, the electronic mask used originally for the 30°V experiments (Figure 3B) was altered to generate a V with a capped end to form a teardrop-shaped enclosed volume (Figure 4, A, C, and D; Supplemental Videos 8 and 9). Confining the aster in this way also let us test a fundamental prediction of the pushing model, that longer MTs on the leading side of the aMTOC would generate less force, due to the length dependence of buckling, than shorter MTs on the lagging side of the aMTOC. To determine whether aMTOCs could center in teardrop enclosures and establish a baseline behavior for coupled aster growth and aster movement, we tracked the movement of aMTOCs from the vertex of the teardrops and plotted aMTOC position as a function of time (Figure 4, C and F). Results from coupled experiments conducted using open 30°V barriers (taken from Figure 3A) are included for comparison (Figure 4, B and E). We observed similar aMTOC movement behaviors in coupled experiments for both the open 30°V (Figure 4, B and E) and the enclosed teardrops (Figure 4, C and F). For the uncoupled condition (Figure 4, D and G), aMTOCs were permitted to nucleate MTs until the network had expanded to fill the interior of the teardrop for ∼15 min, after which the aMTOCs were released and tracked. Immediate movement of the released aMTOCs suggested that they were able to move away from the vertex and through the preexisting MT network, albeit not quite as fast or as far as compared with the other experimental conditions. These observations suggest that the aMTOC can move independently from the bulk MT network. We also observed MT bending in front of the moving aMTOCs (Figure 4, white arrowheads), suggesting that the filaments were under a compressive load and unable to oppose the larger pushing forces generated by shorter growing MTs behind the aMTOC.

FIGURE 4:

Aster movement does not require bulk translocation of the entire MT network. The cartoon in A illustrates the approach used to decouple aster growth and aMTOC movement. aMTOCs were partially embedded in PEGdiPDA posts using the methodology described in Figure 3A such that they could nucleate MTs from their extract-exposed surfaces before being released. Time-lapse spinning-disk confocal microscopy was used to visualize MTs labeled with mCherry-TMBD as in Figure 1. The yellow dashed line at each time point in B extends from the aMTOC position (red asterisk) and is included to facilitate comparison of aMTOC velocities after release at time = 0. The images in B and C show aMTOC movement at the start of nucleation away from the vertex of a 30°V and a 30° teardrop, respectively. In contrast, the images in D show an aMTOC only partially embedded in the PEGdiPDA post and allowed to nucleate MTs for 15 min before light-induced release as described in A. For each experimental condition, aMTOC movement away from the barrier’s vertex was plotted as a function of time with positional data from B, C, and D shown in graphs E, F, and G, respectively. Shaded outlines represent 95% confidence intervals; gray lines represent each experiment; n ≥ 6. Yellow arrowheads point to MT bending. For all images, scale bars = 20 µm.

Movie S8

Aster moving away from the vertex of a teardrop with untreated extract.

Movie S9

Aster nucleating MTs, then being released, and moving away from the vertex of a teardrop with untreated extract.

DISCUSSION

Our results implicate a pushing-based force generation mechanism as playing a larger role in aster movement and centration than previously acknowledged and one that could be the predominant mechanism underlying aster movements within large cells, at least on a scale of ∼100 µm or less in diameter. We found that aMTOC-nucleated asters can indeed find the center of PDMS channels as well as annular cylinder hydrogel enclosures, suggesting that our model system might serve as a reasonable proxy to study the phenomenon. In agreement with recent published studies in sea urchin embryos, (Meaders ), inhibition of dynein via p150-CC1 did not affect aster centration in our extracts. Taken at face value, this would suggest that aster centration, over the length scales tested, can occur independently of dynein function and would implicitly rule out a dynein-dependent pulling-based mechanism. However, this result contradicts observations of the effects of dynein inhibition on aster centration over similar length scales in vivo. Specifically, in large sea urchin blastomeres (∼100 µm in diameter), treatment with the dynein inhibitor ciliobrevin was shown to negatively affect aster centration (Tanimoto ). p150-CC1 injection into X. laevis embryos (∼1.2 mm in diameter) resulted in asters that failed to reach the cell center after traveling ∼200 µm from the blastomere cortex, suggesting that dynein activity is not required for aster movement over shorter length scales but might be required for translocation over longer distances (Wuhr ). Even in Xenopus egg extracts, dynein inhibition (via ciliobrevin and p150-CC1, respectively) affects both MT aster-dependent compartmentalization (Cheng and Ferrell, 2019) and aster–aster separation in bulk X. laevis egg extracts (Pelletier ). In the context of a pushing-versus-pulling argument, we should emphasize that we used PEGDA and PDMS barriers in our in vitro experimental approach. Characterization of the architectures of asters grown on slabs made of these materials and the observed MT slipping as shown in Figure 1 suggests that they are largely inert in terms of motor binding, a conclusion that is also supported by the observation of slipping of MT growing ends along vertical barriers (Supplemental Figures S2 and S4). Thus, it is unlikely that motors anchored at the barriers mediate cortical pulling, though it is abundantly clear that this mechanism is involved in spindle centration in vivo (Aist ; Adames and Cooper, 2000; Labbe ; Park and Rose, 2008; Kotak ; Okumura ). Taken as a whole, these data suggest that it is possible for MT-dependent pushing forces to center asters in cells up to ∼100 µm in diameter.

The observation that aMTOC asters failed to reach the geometric center of our lobster trap enclosures and instead adopted a decentered final position in which the aMTOC was not spatially coincident with the aster CoMM is consistent with a force contribution from a pushing-based mechanism and implicitly argues against a model that relies solely on cytoplasmic pulling, which would predict that the aster would indeed be able reach the geometric center. We propose that this failure to center is caused by the collection and focusing of growing MTs into the triangular recesses on either side of the trap’s constriction and by the resultant generation of pushing forces that work to keep the aster in the lower chamber of the trap. This argument requires that MTs continue to grow once they hit a hydrogel barrier and is supported by our observation that when EB1-decorated growing ends (comets) encounter a hydrogel barrier, they indeed continue to grow and will either remain at the point of impact or slide along the barrier. depending on the angle of contact (Figure 1A; Supplemental Figure S4). The same logic could be applied to explain the results of experiments in sawtooth channels, in which asters reached equilibrium positions farther away, on average, from the sawtooth side of the channel, consistent with the idea that pushing forces against the channel walls were uneven and were made more productive by the sawtooth geometry. We acknowledge that the results from our lobster trap experiments do not definitively rule out the possibility that pulling forces are still at play and that aster centration is prevented in these channels not by a pushing mechanism, but instead by steric constraints imposed by the physical properties of the aster, that is, the aster is simply too large and too rigid to be pulled through the constriction. Recent characterizations of aster dynamics in F-actin–intact egg extracts indeed suggest that these assemblies behave as elastic, deformable gels (Pelletier ). However, we argue that if pulling forces play a predominant role in centration, the dynamic nature of aster MTs should still allow the aster to reorganize around the constriction over time and ultimately reach the center of these devices. Clearly, more work needs to be done to measure the mechanical properties of the aster as a whole.

Our observations of aMTOC movements away from V-shaped hydrogel barriers make a compelling case for a pushing-based mechanism. This geometry provides little opportunity for motors to anchor to the barrier in a way that would allow for the aster behavior that we observed. We are confident that our PEGDA hydrogels are relatively inert and minimize the potential for nonspecific protein binding (Nuttelman ; LeValley ). As the asters always moved out of the V-shaped barriers toward the opening, this suggests that cortical pulling is likely not a major contributor to the phenomenon. The aMTOC velocity was higher and they moved farther from the vertex in 30°V barriers, a geometry that allowed for closer proximity to the wall and potentially larger pushing forces to be generated, relative to similar experiments using 90°V barriers. Additionally, the relationship between velocity and aster asymmetry deviated from what one would expect if a cytoplasmic pulling mechanism was responsible, that is, a positive correlation between velocity and the separation distance between the aMTOC and the aster CoMM, which was up to ∼70 µm with the V-shaped barriers. We instead observed a negative correlation, which we posit is likely the result of a combination of reduced pushing forces due to an increasingly larger separation between the aMTOC and nearest barrier and of a concomitant increase in resistive drag as the aster grows. Finally, we observed aMTOC movement away from barriers even when the aster CoMM was behind the moving aMTOC, which is most consistent with a pushing-based mechanism.

The compressive force that a single MT can bear or exert is thought to fall off as a function of 1/(MT length2) according to the Euler buckling theory (Dogterom and Yurke, 1997; Kimura and Onami, 2005). Studies in vitro and mathematical modeling of single MTs suggest that the maximum force a single MT can exert before buckling is on the order of 3−10 pN and occurs at lengths of ∼15 µm (Dogterom and Yurke, 1997; Holy ; Zhu ), far too short to account for the distances traveled by asters in our system and in the living blastomeres of specific echinoderms (Minc ; Tanimoto ) and amphibians (Wuhr ). It should be noted, however, that studies of MT bending modes in cells (Brangwynne ) suggest that MTs can withstand much larger compressive loads than predicted by in vitro observations. This line of reasoning and the negative effects of ciliobrevin treatment on aster migration (Tanimoto ) have led many researchers to discount pushing as a possible mechanism. We argue that a branched MT network might allow for pushing forces to act over longer distances. Indeed, the relatively constant density of growing MT ends in expanding asters (Ishihara , 2016), coupled with MT-dependent MT nucleation (Petry ), is consistent with the idea that the distance between the aMTOC and a barrier is bridged by a tiled array of interconnected and short MTs, not by long MTs spanning the entire distance (Figure 5). Similarly, MT minus ends are distributed throughout meiotic monopoles in X. laevis extracts (Decker ). An interconnected network of MTs is likely better able to bear a compressive load, but its ability to do so is still dependent on the length over which the compressive load is borne. Indeed, following aster release in teardrop enclosures, the longer MT network on the leading side of the aMTOC seemed prone to buckling, whereas the shorter MT network on the lagging side seemed less so. Regardless, more work has to be done to characterize the changes in MT dynamics and network organization that likely accompany aster movement and centration. Indeed, our observations that aster movement can be decoupled from aster MT growth suggest that some degree of structural independence is present within the aster. Recent experimental evidence suggests that this is indeed the case, as nuclei and associated centrosomes have been shown to center in tessellated arrays of MT asters formed in cycling extracts (Cheng and Ferrell, 2019).

FIGURE 5:

Cartoon summary for how MT-based pushing forces might facilitate aster movement in our system. The cartoon in A depicts how asters might behave assuming an unbranched, radial elongation model of aster growth. Initially, short unbundled MTs extending from the aMTOC and reaching the proximal barrier surface might be able to generate sufficient force to produce aster movement. However, later in time, once MTs approach their critical length, they would buckle under compressive loads, resulting in reduced force generation and a failure to center (dashed yellow line). The cartoon in B shows an architecturally more complicated aster with MT branching and bundling to effectively brace MTs and allow for more growing ends to impinge on the proximal barrier surface. In this model, MTs would be sufficiently buttressed by interactions with other MTs, resulting in a larger pushing force and, ultimately, a longer distance through which asters could traverse.

So, how would we characterize our system, and what exactly is it telling us about aster centration? We like to think that the model system used here, cell-free extracts derived from frog eggs, lies somewhere on the continuum between in vitro and in vivo (but is closer to the in vivo end of the spectrum). Despite the fact that many fundamental mechanisms elucidated using Xenopus egg extracts have proven to be conserved, we acknowledge that this is indeed not an in vivo system. Furthermore, we note that we are also confining extracts and asters within artificial enclosures that most certainly fail to accurately mimic the plasma membrane in terms of its mechanical, biochemical, and geometrical properties. Therefore, many forms of regulation or feedback on MT dynamics that the plasma membrane normally confers in vivo are likely missing from our system. We also acknowledge that egg extract is not a true mitotic system. For all of the aforementioned reasons, we suggest using caution when extending our findings to explain aster and spindle positioning as it occurs in cells in vivo. Indeed, it is well established that cortical pulling forces play a major role in aster/spindle positioning in several eukaryotic systems (for excellent reviews, see Kotak and Gonczy, 2013; McNally, 2013; Meaders and Burgess, 2020). Furthermore, data in support of the cytoplasmic pulling model, which are largely based on observations of aster and spindle positioning in sea urchin and in C. elegans blastomeres, are compelling (Hyman, 1989; Gonczy ; Kimura and Onami, 2005; Kimura and Kimura, 2011; Tanimoto , 2018). For example, in sea urchins, inducing a rapid aster asymmetry via laser ablation of MTs on one side of an aster results in the aster moving away from the site of ablation, an observation consistent with a pulling mechanism but difficult to reconcile with cortical pushing. In these same model systems, inhibition of dynein (via treatment with ciliobrevin; Tanimoto , 2018) inhibits the ability of asters to center, broadly implicating a pulling-based mechanism. Even in X. laevis egg extracts, the centration of individual nuclei—and presumably their associated centrosomes—within each tessellated tile formed by an aster, is abrogated upon treatment with ciliobrevin (Cheng and Ferrell, 2019; Field ). Finally, when comparing our centration dynamics to those that occur in X. laevis embryos, there are clear differences as well. The velocities observed in this system (∼3−5 µm/min) are noticeably lower than those measured in vivo (∼7 µm/min; Wuhr ). Furthermore, aster centration in single-celled X. laevis embryos requires dynein, but in this system the aster center has to travel upward of 600 µm to reach the cell center. Interestingly, even in embryos treated with p150-CC1, asters were able to move at least 200 µm away from the cortex (Wuhr ). This would suggest that there is a limit to how far asters can generate a productive pushing force and that aster movement beyond this point likely requires another mechanism.

In summary, the current state of the field leaves room for debate and, collectively, requires better approaches to provide more definitive answers. We believe that our data, taken as a whole, suggest that a pushing-based force mechanism is capable of moving an aster through cytoplasm and that these forces can act over longer-than-expected distances in the in vitro system used here. Whether the same mechanism functions in large cells in vivo is a different and, at present, unresolved question. However, recent in vivo experimental evidence argues that this might be the case (Meaders ).

MATERIALS AND METHODS

Mold and device manufacturing

Multi-depth, three-dimensional (3D) printed microfluidic positive molds with deep fluid flow stabilization chambers (3648 × 1000 × 600 µm) and shallow fluid channels (2000 × 100 × 30 µm) were fabricated on preacrylated glass slides (microscope slides, 75 × 50 mm; Corning). The 3D printer (MAX X27; Asiga) used a 3D printing method for microfabrication, named stereolithography (SL), which is based on digital light processing (DLP; Lu ; Gross ; Gong ; Ho ). The process to manufacture a 3D multiheight mold is illustrated in Supplemental Figure S6A. First, a preacrylated glass slide was attached to the 3D printer build platform with double-sided tape. UV light was transmitted from the bottom of the resin tray to polymerize the first layer in accordance with the micromirror design (Supplemental Figure S6C). To make the next layer, the 3D printer adjusted its height and exposed the resin to UV light again, building upon the established layer. This continued until the desired mold, which is a positive relief of the microfluidic device, was finished being fabricated (Supplemental Figure S6D). The steps needed to turn the mold into a functional microfluidic device are shown in Supplemental Figure S6B. First, the mold was placed in a developing tank filled with isopropyl alcohol to flush away residual resins. This was followed by 5−10 min of UV curing, and then the mold was treated with oxygen plasma under 600 psi for 5 min to remove any remaining residual resin chemicals on the surface of the micropattern. Next, the mold and a weigh boat containing 50 µl fluorosilan were placed together in a desiccator for 3 h so that a thin layer of fluorosilan was uniformly deposited on the mold’s surface. Once the surface treatment was completed, a PDMS elastomer (Dow Corning) containing a base and curing agent was poured into the mold and allowed to cure at room temperature for 2 d. Once cured, the PDMS replica was cut out and peeled off the resin mold (Supplemental Figure S6E). Holes were punched into the inlets and outlets with a sharpened 20G dispensing needle (Brico Medical Supplies). Finished microfluidic devices were made by bonding the PDMS replica to a coverslip (#1.5 thickness; Fisher Scientific) after oxygen plasma treatment (Duffy ; Supplemental Figure S6F).

Hydrogel polymers

The hydrogel annular cylinders, sawtooth arrays, 30° and 90°Vs, and teardrop structures were polymerized by using 30% (wt/wt) PEGDA in 1× CSF-XB (100 mM K[Cl], 0.1 mM CaCl2, 1 mM MgCl2, 50 mM sucrose, 10 mM HEPES, and 5 mM ethylene glycol tetraacetic acid (EGTA) at pH 7.7) with 1% (wt/wt) photoinitiator (lithium phenyl-2,4,6-trimethylbenzoylphosphinate, LAP). An IX81 Olympus microscope with attached micromirror array system (Polygon400 dense pixel; Mightex) was used in conjunction with an Olympus 20 × UPlanSApo (0.75 NA) air objective to expose hydrogel solutions to polymerizing UV light (405 nm) in microfluidic devices (Hahn ). PEGdiPDA (number average molecular weight ∼4070 Da) was synthesized as described (Kloxin ; Kharkar ) and was mixed with 1× CSF-XB buffer at 10% (wt/wt) with 1% (wt/wt) LAP. This monomer was photopolymerized into a hydrogel polymer matrix by exposure to UV light (405 nm) with the micromirror system described above and was then photodegraded as needed by exposure to a lower wavelength (365 nm) (Fairbanks ).

Aster formation in X. laevis interphase extracts

X. laevis egg extracts were prepared from oocytes arrested in meiotic metaphase (Desai ; Maresca and Heald, 2006; Good and Heald, 2018). To induce interphase for aster formation and to block protein synthesis and prevent reentry into mitosis, 0.4 mM CaCl2 and 100 μg/ml cycloheximide, respectively, were added to the arrested egg extract on ice. The extract was also supplemented with 150 nM mCherry-TMBD to label the MTs. The mCherry-TMBD fusion protein was produced and purified as described (Mooney ). Briefly, the plasmids were expressed in BL21 (DE3) cells, and the resulting fusion protein was purified on nickel spin columns (Bio-Rad). In experiments that inhibited functional dynein, 2 μM of p150-CC1 was added to the extract (Gaetz and Kapoor, 2004). For experiments in which MT growing ends were visualized, 60 nM EB1-GFP (Tirnauer ) was added to the interphase extract. To generate asters, aMTOCs (Tsai and Zheng, 2005), which consist of protein A Dynabeads (Invitrogen) coated in aurora A antibodies (a generous gift from T. Mitchison and C. Field, Harvard Medical School), were added to the egg extract. All experiments using egg extract were repeated at least three times using different extract preparations. All experiments involving X. laevis were conducted in accordance with the University of Wyoming and the Marine Biological Laboratory Institutional Animal Care and Use Committees’ guidelines.

Microfluidic device preparation

To ensure that hydrogel structures adhered to the glass coverslips within our PDMS devices, a 2% (vol/vol) solution of 3-acryloxypropyltrimethoxysilane solution in 95% ethanol was flowed through each device for 5 min followed by a wash with 95% ethanol. Treated devices were subsequently baked at 75°C for 10 min and then washed again with 95% ethanol before being rinsed in ddH2O and baked dry. After hydrogel structures had been made in the microfluidic devices, the glass on the interior of the microfluidic devices was passivated by incubation with 0.5% (wt/vol) pll(50)(HCl)-g(10)-mPEG(114) (pll-PEG; Nanosoft Polymers) in 1× CSF-XB buffer for at least 1 h at room temperature. For experiments using exposed aMTOCs, coverslips were passivated by exposing a device filled with poly(ethylene glycol)mono acrylate to UV light. This treatment results in a brush layer of PEG on the coverslip surface that hydrogel structures made from PEGDA and PEG-diPDA can adhere to. Afterward the remaining PEG-monoacrylate was washed out. Following this passivation, the devices were stored in 50-ml conical tubes containing 1× CSF-XB buffer overnight at room temperature to ensure that the device was well saturated to prevent fluid absorption and flow. Before the microfluidic device could be used, it had to be washed with ≥100 μl of 1× CSF-XB buffer. A 1-ml syringe (Norm-ject) connected to 0.10-in. tubing (Tygon) was filled with 1× CSF-XB buffer, and the tubing was inserted into the inlet to pump ≥100 channel volumes of buffer through the device. While the device was being washed, the egg extract was taken up into a fresh syringe and tubing. The tubing was inserted into the inlet of the microfluidic device, and the egg extract was pumped through at 5 μl/min using a syringe pump (Nemesys pumps; CETONI GmbH, Korbussen, Germany) in a 4°C room for 20 min. The device was brought to the cooled microscope room (16°C) on ice to ensure that MT polymerization did not occur prematurely. While still being kept cold, the glass coverslip was affixed to the cooling unit with VALAP (made by melting together equal weights of Vaseline, lanolin, and paraffin). The cooling unit consisted of a hollow, custom-fabricated metal rectangle through which ice water was continuously pumped. This maintained a steady temperature on the microscope stage over long time periods.

Imaging and analyses

Imaging was done on an Olympus IX81 confocal microscope equipped with a spinning-disk confocal head (CSU-X1; Yokogawa). Olympus 20 × UPlansSApo (0.85 NA) oil and 40 × UPlanFLn (1.3 NA) oil objectives were used to visualize MTs. An Olympus 60 × UPlanSApo (1.35 NA) oil objective was used to visualize EB1 comets interacting with the hydrogel barrier. For the channel experiments, 25-µm z stacks (taken at 5 µm intervals) were generated for each of the positions every 5 min and the plane with a visible aMTOC was used. For experiments using hydrogel lobster traps and V-shaped barriers, single-plane images were taken every minute and asters were tracked only after their MTs made contact with the nearest wall. The same was done for the annular cylinders, but images were taken every 30 s. Images were processed and analyzed using Fiji (National Institutes of Health; http://rsb.info.nih.gov/ij/) to manually get the coordinates of the aMTOC to track aster movement. Images were background subtracted based on the signal seen in the hydrogel structures alone and cropped. The enhanced contrast feature in Fiji (saturated pixel 0.4%) was used on the time-lapse videos to consistently adjust the LUT for aster images in all of the montages. Aster motion was calculated based on the aMTOC coordinates measured in Fiji, which were imported into GraphPad Prism (version 8; GraphPad Software) to make both the series of line traces and the line graphs with 95% confidence intervals. The CoMM was calculated using Fiji. First the images were background subtracted to remove signal from extract devoid of MTs. Images were then smoothed using a rolling ball filter (radius = 5 pixels) and segmented based on the MT signal such that the segmentation resulted in a single, contiguous region of interest (ROI) for MTs associated with the aster. The Center of Mass function of the measure tool in Fiji was then applied and produced the coordinates of the ROI’s center of mass. Then the aMTOC coordinates were subtracted from the CoMM to find the difference. The velocities to be graphed relative to this difference and to the proximity to the nearest barrier were found by taking the slope at 5-min intervals from the 30°V and 90°V plots in Figure 3D.

Click here for additional data file.

Click here for additional data file.