Microtubule Growth Rates Are Sensitive to Global and Local Changes in Microtubule Plus-End Density: Current Biology | Current Biology
Zachary M. Geisterfer, Daniel Y. Zhu, Timothy J. Mitchison, John Oakey, Jesse C. Gatlin
Results and Discussion
Microtubule-based assemblies, specifically interphase microtubule asters and mitotic spindles, have been characterized in discrete droplets of cell-free Xenopus egg extract to study microtubule self-organization and scaling phenomena [
Effects of confinement on the self-organization of microtubules and motors.
]. In these emulsion droplets, however, it is difficult to collect images with a high enough signal-to-noise ratio and sufficient temporal resolution to characterize dynamic molecular-scale phenomena such as microtubule growth. To circumvent these limitations in a way that would still allow for precise control of extract volume, we used hydrogel photolithography [
Fabrication of poly(ethylene glycol) hydrogel microstructures using photolithography.
] to confine cell-free extracts and microtubule asters within microscale enclosures of precise geometrical shape and size (Figure 1A).
To generate photo-patterned structures on the coverslip surface, we placed a digital micro-mirror array in the optical path of a microscope and projected light patterns onto a pre-polymer solution of poly(ethylene glycol) diacrylate (PEGDA) contained within a polydimethylsiloxane (PDMS) microfluidic flow chamber 30 μm in height (Figures 1A and 1B; leftmost and center-left panels; Figure S1A). The positions of these enclosures within the device were dictated by the random spatial arrangement of artificial microtubule organizing centers (aMTOCs; [
Aurora A kinase-coated beads function as microtubule-organizing centers and enhance RanGTP-induced spindle assembly.
]) within the pre-polymer solution (Video S1). By first generating a flow of extract into the device from one direction to fill the enclosures (Figures 1A and 1B; center-right panel), and then subsequently flowing in an oil phase (fluorinated oil/surfactant) from the opposite direction (Figures 1A and 1B; rightmost panel), we could trap aqueous volumes of extract within the confines of the hydrogel micro-enclosures (Figures 1A and 1B). It should be noted that the exterior tear-drop shape of the enclosures was important for isolating extract (or other aqueous phases) in hydrogel structures using oil crossflow (Figure S1B). With this experimental paradigm, we successfully isolated discrete volumes of cell-free extracts at the coverslip surface (Figures 1A and 1B, right-most panel) with precise control of geometry and volume. aMTOCs confined in this manner were able to nucleate microtubules and generate microtubule asters as described previously in bulk interphase egg extracts [
Evidence for an upper limit to mitotic spindle length.
], we predicted that changes in cytoplasmic volume might affect microtubule growth rates. To test this hypothesis, we used our hydrogel experimental platform to capture aMTOCs in cylindrical micro-enclosures of increasing diameter. The range of diameters tested (equivalent to spherical cells ranging from 30 to 115 μm in diameter [Ø]) was chosen based on the range of blastomere sizes in which in vivo mitotic spindle scaling has been observed ([
Evidence for an upper limit to mitotic spindle length.
]; Figure 2A). The extracts used in these experiments were supplemented with a low concentration of EB1-GFP (60 nM) to visualize and track growing microtubule plus ends. Subsequent analysis of time-lapse recordings of EB1 comets using both automated tracking and manual kymographs allowed us to measure microtubule growth rates and plot them versus micro-enclosure volume (Figures 2B and S1E).
Microtubule growth rates showed some correlation with cytoplasmic volume within our system; however, most of the observed variation could not be explained by a linear model (R2 = 0.45). For example, in volumes as low as ∼13 pL, we observed growth rates that were statistically higher or indistinguishable from those found in ∼160-pL enclosures (corresponding to spherical cells of ∼30 and 65 μm Ø, respectively), and microtubule growth rates seemed to plateau at volumes above ∼400 pL (spherical cells of ∼95 μm Ø; Figure 2B). We also observed large, statistically significant differences in microtubule growth rates under near-isovolumetric conditions (e.g., see the spread observed in the data points for enclosures around 400 pL in volume in Figures 2B and S1D). Taken as a whole, these observations prompted us to identify other factors that might better account for the observed variation of microtubule growth rates. Indeed, when we compared EB1-GFP signal densities within each ∼400 pL device, we found large differences in the total number of tracked microtubule growing ends (Figure 2C; Video S2). In contrast, both EB1 signal density and microtubule growth rates observed in ∼160-pL devices (cell diameter of ∼65 μm) showed less device-to-device variation (Figures 2C and S1C; Video S3). We speculated that the measured differences in microtubule growth rates observed in similar cytoplasmic volumes might be caused by differences in microtubule plus-end density, rather than changes in cytoplasmic volume. This would be consistent with a scenario in which each growing microtubule plus end acts as a sink for structural components (i.e., tubulin) and/or +TIP proteins [
Video S3. EB1 Comets in ∼160-pL Micro-enclosures, Related to Figure 2
To determine whether microtubule plus-end density is a more reliable predictor of microtubule growth rates within our system, we re-plotted microtubule growth rates versus EB1 density (Figure 3A). Over the range of EB1 comet densities measured, microtubule plus-end density negatively correlated with microtubule growth rates and better accounted for the device-to-device variation found in microtubule growth rates under near-isovolumetric conditions (Figure 3A; R2 = 0.88. This observation was confirmed using manually recorded growth rates and EB1 densities, suggesting this trend is not an artifact of automated tracking (Figure S1F). These data are consistent with the idea that growth rates are regulated by growing-end competition for a limited supply of components or local regulators [
By this logic, we hypothesized that changes in microtubule plus-end density would lead to correlative changes in microtubule growth rates. To test this, we increased microtubule plus-end density by capturing two aMTOCs in ∼160-pL micro-enclosures (Figure 3B; Video S4). With the additional aMTOC, we indeed observed greater densities of EB1 comets, which resulted in measured decreases in microtubule growth rates (Figure 3A; “2 aMTOC”). Increased EB1 comet density resulting from the additional aMTOC suggests that the nucleation capacity of the extract is not taxed by a single aMTOC. However, analysis of nucleation rates from single aMTOCs suggests that volume-dependent changes in aMTOC nucleation rates plateau at volumes above ∼130 pL (Figure 1G). Overall, these observations confirm that microtubule growth rates are sensitive to changes in global microtubule plus-end density. Importantly, we did not observe the same trend between microtubule plus-end density and catastrophe frequency, which was relatively constant across all extract volumes tested (Figure S2A; R2 = 0.23). Though we acknowledge that changing the number of aMTOC microtubule-nucleation sights might affect microtubule growth rates via a different mechanism (see Discussion), we reasoned that the link between microtubule plus-end density and microtubule growth rates might depend on the availability of key MAPs and their diffusion and capture at microtubule growing ends [
Video S4. Micro-enclosure Featuring an Additional aMTOC, Related to Figures 3 and 4
If microtubule growth rates are diffusion-limited in our system, one would expect spatial differences in microtubule plus-end densities to impart a local effect on microtubule growth rates. To test this, we generated hourglass-shaped micro-enclosures and trapped a single aMTOCs in one of the two connected lobes (Figure 3C). In each of these experiments, the lobe containing the aMTOC exhibited a locally higher EB1 comet density than the unoccupied lobe. A comparison of microtubule growth rates in each lobe showed significant differences in microtubule growth rates (Figure 3D), despite all growing ends being contained within the same continuous cytoplasm. Consistent with our previous findings, the regions containing lower microtubule plus-end densities displayed the highest microtubule growth rates (Figure 3D). This result suggested a possible role for a diffusion-limited mechanism in regulating microtubule growth rates [
Macromolecular crowding pushes catalyzed microtubule growth to near the theoretical limit.
The observed variation in microtubule growth rates within a single cytoplasm led us to explore the length scales over which microtubule growth rates might be sensitive to spatial differences in microtubule plus-end density. To identify the local plus-end density experienced by each EB1 comet over its lifetime, we developed a custom MATLAB code that uses positional data obtained from u-track software [
Robust single-particle tracking in live-cell time-lapse sequences.
]. This code allowed us to define a search area projected from the center of each comet and then to identify and count the number of neighboring EB1 comets found within that area over the lifetime of the tracked comet (Figure 4A). This analysis, termed “local plus-end density,” was then repeated for all tracked EB1 comets in the image series.
We first analyzed the relationship between local plus-end density and microtubule growth rates in a time-lapse image series in which spatial variation in growing plus-end density was qualitatively evident (one of the “two aMTOCs” time series in Figures 3A and 3B; see also Video S4). This allowed us to compare regions of varying plus-end density at similar radial distances from the aMTOC. Moreover, the simple cylindrical geometry of the extract as confined in this device allowed us to rule out the possibility that spatial signaling gradients emanating from the aMTOC might contribute to the differences in microtubule growth rates observed in our two-lobed devices (Figures 3C and 3D). After determining the local plus-end density experienced by each EB1 comet within a 3-μm search radius, we plotted the average local density and average growth rate for each EB1 comet over its lifetime against the x-y position (Figure 4B; center and right panel, respectively). Consistent with the negative correlation, we observed between global plus-end density and mean microtubule growth rate (Figure 2A), regions with lower microtubule growth rates showed higher local microtubule plus-end densities (Figure 4B). Comparisons of EB1 comet speeds using kymographs generated from two regions with distinct EB1 comet densities within the same micro-enclosure (Figure S2E) confirmed a similar relationship.
To better characterize the relationship between local microtubule plus-end density and microtubule growth rates, we performed the local density analysis over a range of different search radii (Figure 4C). The output from each search radius was binned and plotted, with each bin centered on the average local density contained within that bin (Figure 2D). This analysis revealed that local microtubule growth rates were negatively correlated with local microtubule plus-end densities (Figure 4C). Not surprisingly, at smaller search radii, microtubule growth rates were more sensitive to local growing-end density (reflected in the larger negative slopes of linear fits). In contrast, as the search radius was increased, the slope of the linear fits approached zero, with the average velocity of each bin closer to the mean global microtubule growth rate of the entire micro-enclosure (red dashed lines, Figure 4C). This relationship between local microtubule plus-end density and microtubule growth rates was also observed across a range of cytoplasmic volumes (Figures S3 and S4), suggesting that diffusion of components, rather than the absolute protein content of the cytoplasm, is responsible for the observed local density effect. In addition, these measurements indicate that the ability of a growing microtubule end to “sense” differences in growing microtubule plus-end density is most acute at distances shorter than a few micrometers. Though we repeated these analyses using smaller search radii, the small sample sizes that resulted precluded a rigorous statistical analysis of the data (data not shown).
Mechanistically, these results suggest (1) that each microtubule growing end might act as a local sink for either tubulin or key regulators of microtubule growth (microtubule plus-end competition), or (2) that steric hindrance and changes in viscosity within dense microtubule polymer networks impede the loading of components onto microtubule plus ends [
Excluded volume as a determinant of macromolecular structure and reactivity.
]. Mechanism (1) requires either a limited source of structural components or the limited translation or rotational diffusion of these same key elements. Predictions for mechanism (2) are less clear, as crowding effects due to large crowding agents (e.g., BSA, PEG, etc.) typically increase rates of chemical reactions through an “excluded volume effect,” whereas small crowding agents (e.g., ethylene glycol, glycerol) slow down the same reactions in a diffusion-limited manner [
Elongation of actin filaments is a diffusion-limited reaction at the barbed end and is accelerated by inert macromolecules.
In summary, our local density analyses suggest that a steady-state dynamic microtubule assembly can experience a global depletion of components, whereas an individual microtubule plus end might experience a local component gradient dictated by the presence and proximity of other growing microtubule ends. We speculate that these density-dependent effects likely exist because of constraints imposed by the slow diffusion of microtubule structural components and/or +TIP-localized growth-promoting factors. The case could be made for free α/β-tubulin heterodimers (tubulin), but a steady-state mean-field model for the concentration of tubulin near a growing microtubule end predicted that local tubulin concentration returned to the bulk concentration at ∼50 nm from the growing microtubule tip [
Estimation of the diffusion-limited rate of microtubule assembly.
], a distance well below the measured length scale of the density-dependent effect we have observed. We note that the in extract conditions used in our studies differ considerably from those assumed by Odde [
Microtubule assembly in clarified Xenopus egg extracts.
]). We acknowledge, however, that it is unclear how much of the total tubulin in a cell is incorporated into aster MTs at steady state and that differences in the ratio of cytoplasmic volume to the number of aMTOC nucleating sites might affect the steady-state partitioning of tubulin in unpredictable ways. Indeed, several labs have characterized a negative correlation between centrosomal nucleation rates and microtubule growth rates [
Divergent microtubule assembly rates after short- versus long-term loss of end-modulating kinesins.
Rather than implicating tubulin as the responsible limiting component, we favor a model that implicates local depletion of some larger, and less abundant, microtubule growth regulator as the mechanistic link between microtubule growing-end density and growth rate. A putative candidate is the processive microtubule polymerase XMAP215, whose activity is known to increase microtubule growth rates in vitro [
Increased microtubule assembly rates influence chromosomal instability in colorectal cancer cells.
], which mirrors reductions in growth rates observed in the EB1 dense regions of our micro-enclosures (see Figure 4C; 2-μm search radius). Furthermore, biochemical characterizations and physical measurements indicate XMAP215 is a large, highly elongated molecule in solution (∼220 kDa and approximately 60 nm in length, and 3.2 nm wide [
XMAP215 activity sets spindle length by controlling the total mass of spindle microtubules.
]. In the absence of concrete measurements, we can only speculate, however, that expected differences in diffusion coefficients and known differences in relative concentrations of tubulin and XMAP215 are sufficient to account for the disparate length scales predicted by theory and those observed here.
The observation that microtubule dynamics are affected by local growing-end density might explain, in part, the variability of published microtubule growth rates, even those measured within the same organism and during similar stages of the cell cycle. In Xenopus egg extracts, interphase microtubule growth rates range from as low as ∼7 μm/min to as high as ∼30 μm/min [
Microtubules growth rate alteration in human endothelial cells.
], respectively. This variability in measured microtubule growth rates far exceeds the ∼2% variability that would be expected from a simple Poisson model of tubulin addition to the microtubule plus end [
Concentration dependence of variability in growth rates of microtubules.
]. Explanations for this variation have been numerous and experimentally intractable, as many reasonable causes have been evoked, e.g., proximity to membranes, spatial effects resulting from changes in MAP function and post-translational modifications, and steric crowding effects [
Tracking of plus-ends reveals microtubule functional diversity in different cell types.
]. Our results suggest that some of this variability can be accounted for simply by spatial differences in growing microtubule plus-end density.
In summary, our observations suggest that microtubule growth rates are regulated by the presence and proximity of other microtubule plus ends and that this spatial regulation can impart local changes in the dynamics of microtubule subpopulations within a single, continuous cytoplasm. This might explain recently observed differences in astral microtubule and spindle microtubule growth rates in C. elegans embryos [
Microtubule dynamics scale with cell size to set spindle length and assembly timing.
]. We postulate that this mechanism might also be biologically significant in several additional contexts, such as mitotic spindle size scaling during development, as relatively small changes in microtubule growth rates have been shown to correlate with large changes in mitotic spindle size [