Microtubule dynamics regulation contributes to endothelial morphogenesis.

Because little is known how microtubules contribute to cell migration in a physiological three-dimensional environment, we analyzed microtubule function and dynamics during in vitro angiogenesis in which endothelial cells form networks on a reconstituted basement membrane. Endothelial network formation resulted from distinct cell behaviors: matrix reorganization by myosin-mediated contractile forces, and active cell migration along reorganized, bundled matrix fibers. Inhibition of microtubule dynamics inhibited persistent cell migration, but not matrix reorganization. In addition, microtubule polymerization dynamics and CLASP2-binding to microtubules were spatially regulated to promote microtubule growth into endothelial cell protrusions along matrix tension tracks. We propose that microtubules counter-act contractile forces of the cortical actin cytoskeleton and are required to stabilize endothelial cell protrusions in a soft three-dimensional environment.

Introduction

The basic mechanism of adherent cell migration in a two-dimensional (2D) environment has been studied extensively, and is thought to predominantly rely on a coordinated cycle of actin polymerization-mediated protrusion, integrin-mediated adhesion and cell body translocation resulting from actomyosin contractility. However, in most cells directional migration also requires an intact microtubule cytoskeleton. Microtubules are rigid cytoskeletal polymers of α/β-tubulin dimers that stochastically switch between phases of growth and shortening, a non-equilibrium polymerization behavior referred to as dynamic instability. In many cell types such as fibroblasts and epithelial cells migrating on a 2D surface, microtubules are preferentially oriented toward the front of the cell through regulated interactions with the cortical cytoskeleton. Such microtubule interactions with the cell cortex are likely mediated by a class of proteins that specifically associate with growing microtubule ends, and may stabilize microtubules by forming dynamic connections with the actin cytoskeleton or near adhesion sites.- Cells migrating in 2D lose directional polarity and cease migration when microtubule dynamics are inhibited, possibly due to disrupted Rho GTPase signaling networks required for establishment or maintenance of directional polarity.,-

In a physiological three-dimensional (3D) environment, however, cell shape changes and interactions with the extracellular matrix are considerably more complex. Migrating cells may employ additional alternative modes of translocation, and the function and regulation of the microtubule cytoskeleton during cell migration in a 3D extracellular matrix is incompletely understood. Cells squeezing through a meshwork of extracellular matrix may utilize a more amoeboid form of migration during which microtubule organization is entirely different, and frequently adapt collective modes of migration, for example during the sprouting of new blood vessels during angiogenesis. To analyze the role of dynamic microtubules in a more physiological cell migration model, we employed a widely used in vitro angiogenesis assay. Primary endothelial cells grown on top of a compliant reconstituted basement membrane matrix reorganize into a capillary-like network within hours. Because cells are not suspended within the Matrigel matrix, this system is a hybrid between 2D and 3D. Nevertheless, endothelial cells experience a pliable, physiological environment, which recapitulates many aspects of angiogenesis in vivo., Here, we propose that spatial regulation of microtubule dynamics promotes the formation of microtubule bundles in endothelial cell protrusions along tension-induced aligned matrix fibers, and that these microtubule bundles are important for persistent migration by counter-acting myosin-mediated contractile forces.

Results and Discussion

Endothelial cell migration along matrix tension tracks requires dynamic microtubules

To analyze endothelial network formation dynamics, we observed HUVEC primary endothelial cells cultured on a reconstituted Matrigel basement membrane by phase contrast time-lapse microscopy for up to 12 h. In control HUVECs, different cell behaviors contributed to endothelial network formation (Fig. 1A; Vid. S1). Cells adhered and spread on the Matrigel substrate within ~30 min after plating. After 2–3 h, most cells had formed small clusters, and contractile forces between cell clusters resulted in reorganization of the extracellular matrix and in the alignment of matrix fibers. These tension tracks were clearly visible by phase contrast microscopy, and typically formed between two opposing cells (Fig. 1A and D; Vid. S1). Individual cells responded to this mechanical stimulus by forming long thin protrusions along such bundles of aligned matrix fibers (Fig. 1D).

Figure 1. Endothelial network formation requires dynamic microtubules and myosin-mediated contractility. Individual images from time-lapse sequences of network formation of control HUVEC primary endothelial cells (A), and cells in the presence of 100 nM nocodazole (B) or 25 µM blebbistatin (C). Scale bar, 200 µm. Arrowheads in (A), 4 h, indicate tension tracks in the control. Insets in the 12 h time-point are higher magnification. The control inset shows a cord of endothelial cells, the nocodazole inset shows a remaining tension track. Each image is stitched from 36 individual images. (D) Magnification of the image region outlined by the dashed box in (A), 4 h, illustrating extension formation of the two opposing cells along a tension track in the matrix. Arrowheads indicate the tips of the cell protrusions. Time stamp is hours and minutes. (E) Quantification of the density of tension tracks 4 h after cell plating on Matrigel in the presence of increasing concentrations of nocodazole or blebbistatin. P values of relevant comparisons are indicated on the graph. (F) Tension track density as a function of time during endothelial network formation in control and 100 nM nocodazole-treated cells. Error bars in (E and F) indicate standard deviation from three independent experiments.

DIC microscopy at higher spatial and temporal resolution around 4–6 h after plating when cells were most migratory confirmed that protrusions resulted from active cell migration along matrix fiber bundles, and were not simply the result of cell stretching by opposing pulling forces on the matrix (Fig. 2A). Cells repeatedly elongated and retracted along these aligned matrix fibers indicating that the generation of matrix tension tracks and subsequent migration along these tracks represent distinct cell behaviors contributing to endothelial network formation. Eventually, cell migration along matrix tracks resulted in cell-cell contact formation of opposing cells to form a thin endothelial cord. Depending on cell density and network topology, these initial cords were either enforced by the migration of additional cells along the cord or disintegrated as a result of collective pulling forces on the matrix. This network consolidation coincided with a gradual disappearance of matrix tension tracks at later time points (Fig. 1F).

Figure 2. Endothelial cell migration along matrix tension tracks. Images from DIC time-lapse sequences of protrusion dynamics in control HUVECs (A), and cells in the presence of 100 nM nocodazole (B) or 25 µM blebbistatin (C). Scale bar, 10 µm. Note the absence of aligned matrix fibers in blebbistatin-treated cells. (D) Representative displacement traces of the protrusion tip as a function of time of six different control, nocodazole-, and blebbistatin-treated cells each, indicating that protrusions collapse more frequently in nocodazole-treated cells. (E) Quantification of the length of protrusion phases along matrix tension tracks in control and 100 nM nocodazole-treated cells from eight 45 min time-lapse sequences per condition. Each circle represents a protrusion phase. Note that because protrusion phases are shorter in nocodazole-treated cells, a larger number of events is measured in the same time period.

As expected, matrix remodeling depended on actomyosin contractility. Myosin II family motors localize to the endothelial cortical cytoskeleton and blebbistatin, a myosin II inhibitor, largely abolished tension track formation (Fig. 1E)., As a result blebbistatin-treated cells remained dispersed and failed to organize into a cell network (Fig. 1C; Vid. S3). However, blebbistatin-treated cells remained motile and formed many, highly branched, randomly oriented protrusions. Although these cells produced abundant lamellipodia, cell protrusions did not continue to extend in the absence of tension tracks (Fig. 2C).

To investigate the role of dynamic microtubules during endothelial network formation, we treated HUVEC cells with a range of low nocodazole concentrations. 100 nM nocodazole did not depolymerize microtubules, but effectively stopped intracellular microtubule polymerization as judged by the disappearance of EB1-EGFP comets, a marker of growing microtubule ends (Fig. 3A)., Yet, 100 nM nocodazole severely disrupted endothelial network formation (Fig. 1B; Vid. S2). As expected, cells became more contractile and the number of matrix tension tracks increased significantly in 100 nM nocodazole-treated cells (Fig. 1E). However, cells failed to productively migrate along these tracks, which resulted in more tension tracks remaining in nocodazole-treated cultures at later time points (Fig. 1B and F).

Figure 3. Microtubule organization and dynamics in endothelial cells migrating along tension tracks. (A) EB1-EGFP localization in the presence of low concentrations of nocodazole in endothelial cells during network formation. Growing microtubule ends are still visible at 50 nM nocodazole, but are completely absent at 100 nM. (B) F-actin and microtubule immunofluorescence of opposing endothelial cell protrusions along tension tracks in control and nocodazole-treated cells. (C) Image from a time-lapse sequence of a cell expressing EB1-EGFP and mCherry-actin. Insets show individual channels. (D) Kymograph of EB1-EGFP in the thin distal part of the protrusion of the cell shown in (C) demonstrating persistent microtubule growth toward the distal tip of the protrusion. The asterisk indicates deformation of the distal protrusion tip by a growing microtubule.

Analysis of DIC microscopy time-lapse sequences further demonstrated that although nocodazole-treated cells were still protrusive, the persistency of protrusions along tension tracks was substantially decreased (Fig. 2D and E). The average life time of a protrusion phase in 100 nM nocodazole was decreased about 3-fold as compared with control cells. Although it has been shown previously that microtubule depolymerization or inhibition of microtubule dynamics disrupts endothelial network formation,,, our results demonstrate that dynamic remodeling of the microtubule network is specifically required for persistent cell migration along aligned matrix fibers, but not for the generation of endothelial cell pulling forces on the matrix.

Microtubule dynamics are spatially regulated in endothelial cells migrating along matrix tension tracks

To analyze how microtubules contributed to protrusion persistency, we first fixed and stained F-actin and microtubules in control and nocodazole-treated cells 4–6 h after plating on Matrigel (Fig. 3B). Control cells stretched along matrix tension tracks and always contained abundant parallel microtubule arrays that typically extended into the very tip of the protrusion. Cells treated with 100 nM nocodazole still contained numerous microtubules in the cell body confirming that this low nocodazole concentration was not sufficient to completely depolymerize microtubules even after prolonged incubation. However, microtubules in nocodazole-treated cells appeared disorganized and did not extend into cell protrusions (Fig. 3B). This indicated that microtubule polymerization is regulated differently in the cell body compared with protrusions along tension tracks.

To test this directly, we analyzed microtubule dynamics by spinning disk confocal live cell microscopy. We used EB1-EGFP as a marker of growing microtubule plus ends because microtubules were too dense in these protrusions to directly observe dynamic microtubules. Dual-wavelength imaging in combination with mCherry-actin revealed that microtubules actively grew into HUVEC cell protrusions (Fig. 3C). Although EB1-EGFP-labeled growing microtubule ends were distributed throughout the cell, the direction of microtubule growth in protrusions along tension tracks was almost exclusively toward the distal end of the protrusion (Figs. 3D and 4A; Vid. S4). Occasionally, we observed growing microtubule ends that bumped into and deformed the membrane at the distal tip of the protrusion (Fig. 3D). Because microtubules have a high bending rigidity, this suggests that microtubules may mechanically contribute to protrusion stability in endothelial cells migrating along matrix fiber tracks in a soft 3D environment.

Figure 4. Spatiotemporal microtubule regulation during endothelial cell network formation. (A) Image of a time-lapse sequence of an EB1-EGFP expressing cell acquired at 2 frames per second. EB1-EGFP positive, growing microtubule ends are distributed throughout the cell. Insets show the regions in the cell body and the protrusion indicated by dashed boxes at higher magnification. (B) Box-and-whisker plot of microtubule growth rates analyzed by EB1-EGFP comet tracking in the cell body compared with the cell protrusion. Outliers are indicated by circles. Notches indicate 95% confidence intervals. n = 300 microtubules. (C) Distribution of wild-type EGFP-CLASP2 or non-phosphorylatable EGFP-CLASP2(9 × S/A) in which GSK3β phosphorylation sites are replaced with alanines. This indicates spatial regulation of wild-type EGFP-CLASP2 in endothelial cells in a 3D matrix. Insets show the regions in the cell body and the protrusion indicated by dashed boxes at higher magnification. Scale bars, 10 µm.

To determine whether microtubule dynamics were spatially regulated in endothelial cells migrating along tension tracks we employed automated tracking of EB1-EGFP comets in time-lapse sequences acquired at high temporal resolution (Fig. 4A; Vid. S4)., This analysis was technically challenging due to optical aberrations resulting from the thick Matrigel layer, and the majority of cells not being parallel to the focal plane. Nevertheless, we consistently observed that the microtubule growth rate was significantly higher in endothelial cell protrusions along tension tracks compared with the cell body (Fig. 4B). Although such differences could in principle result from different microtubule network geometries in cell body and protrusion, our recent, detailed analysis of microtubule dynamics in epithelial cells in 3D tissue culture indicated that this is not the case. Analysis of the correlation between EB1-EGFP track length and measured growth rate demonstrated that although microtubule network geometry contributes to the spread of measured growth rates, it does not account for systematic differences in different intracellular domains. In addition, we did not observe a measurable growth rate increase in growth factor-induced protrusions in epithelial cells that appear morphologically quite similar to endothelial cells migrating along tension tracks. In this regard it is interesting to note that low doses of microtubule perturbing drugs increase microtubule dynamics in endothelial cells,,, an effect that also is unique to endothelial cells. Together, these observations suggest endothelial cell-specific mechanisms of microtubule regulation, and it will be interesting to see whether this can be therapeutically exploited.

We and others have shown that CLASPs are microtubule plus-end-tracking proteins that are partially responsible for microtubule interactions with the cortical cytoskeleton in migrating cells.,- In addition, microtubule association of CLASP proteins is spatiotemporally controlled in migrating cells, and dephosphorylation of CLASPs near the leading edge locally increases CLASP-microtubule binding likely as a result of local GSK3β inactivation. To test whether such a gradient of CLASP-microtubule binding also existed in endothelial cells migrating along tension tracks, we analyzed the localization of EGFP-CLASP2. Indeed, although EGFP-CLASP2 evidently bound along microtubule ends in the protrusion, in contrast to EB1-EGFP, EGFP-CLASP2 was completely absent from microtubules in the cell body (Fig. 4C). The bright perinuclear signal is Golgi-associated EGFP-CLASP2 that occurs independent of microtubules. EGFP-CLASP2(9 × S/A) in which all GSK3β phosphorylation sites were replaced with non-phosphorylatable alanine residues associated with microtubule ends equally well in the cell body and the protrusion, confirming that the spatial gradient of CLASP2 microtubule-binding resulted from differences in the GSK3β-mediated CLASP2 phosphorylation state.

In conclusion, our data support a mechanism in which the different dynamic and mechanical properties of the microtubule and actin cytoskeletons cooperate during persistent endothelial cell migration along matrix tracks in a physiological environment. We propose that spatiotemporal regulation of microtubule dynamics results in active microtubule growth into elongating protrusions, and stabilization of these microtubules through interactions of distal microtubule parts with the cortical cytoskeleton near the protrusion tip. This is likely mediated by microtubule plus end-tracking proteins that bind to growing microtubule ends through interactions with EB1., Because adhesion as well as the resistance to contractile forces of a soft 3D extracellular matrix is probably small compared with adhesion on a 2D rigid coverslip, microtubules may play a more important role in 3D. Microtubules crosslinked to the cell cortex may be required to counteract actomyosin contractility and prevent protrusion collapse. This is supported by the observation that remaining microtubules in protrusions of nocodazole-treated cells often appeared buckled (Fig. 3B), which indicates the presence of compressive forces on microtubules as a result of protrusion retraction. This is consistent with a tensegrity model for endothelial cell migration along matrix fiber tracks in 3D in which rigid microtubule bundles act as compression elements that counteract actomyosin-mediated contractile forces., In addition, microtubule growth into long endothelial cell protrusions along matrix tension tracks is likely essential for organelle redistribution and forward transport, and may be required for contact formation during endothelial cord formation.

Materials and Methods

Cell culture

Pooled HUVECs (Invitrogen, Cascade Biologics) were cultured in Medium 200 and Low Serum Growth Supplement (Invitrogen, Cascade Biologics) at 37°C and 5% CO2. For 3D angiogenesis assays, a thin layer of BD Matrix Growth Factor Reduced Matrigel (BD Biosciences) was spread onto the bottom surface of either a 2-well #1 Lab-Tek chambered coverglass (Nalgene Nunc International) or a 35 mm #1 glass bottom dish (MaTek Corporation). The Matrigel was allowed to solidify for ~15 min at 37°C, 5% CO2 before plating ~20,000 cells/cm2. After the cells adhered (~30 min after plating), medium and unattached cells were aspirated and replaced with fresh medium, containing 25 mM HEPES (Gibco) and indicated concentrations of nocodazole (Sigma-Aldrich), (-)-blebbistatin (Sigma-Aldrich) or DMSO as a vehicle control. For live cell imaging, chambers were sealed with silicone vacuum grease (Dow Corning) to prevent media evaporation.

Adenoviral vectors and transduction

mCherry-actin, EB1-EGFP and EGFP-CLASP2 expressing adenovirus particles were produced using the AdEasy system (Agilent Technologies) as described. HUVECs were infected with recombinant adenoviruses for 1 h and allowed to express proteins for 18 h before plating on Matrigel. For double infections, both viruses were added simultaneously.

Immunofluorescence

HUVECs in 3D cultures were washed in warm PBS, briefly fixed in 0.25% glutaraldehyde in BRB80 (80 mM K-PIPES, 1 mM EGTA, 1 mM MgCl2) and then fixed for 5 min in the same fixative also containing 0.1% Triton X-100. Glutaraldehyde autofluorescence was quenched in 0.2% sodium borohydride in PBS. Cells were stained using standard protocols with a rat anti-tubulin antibody (MCA77G, Serotec) and fluorescently labeled phalloidin (Invitrogen) and stored in PBS containing 0.1% sodium azide.

Image acquisition and analysis

EB1-EGFP, EGFP-CLASP2 and mCherry-actin dynamics were imaged at 37°C with a 60X NA 1.20 (Plan Apo VC; Nikon) water immersion objective lens on an inverted microscope system (TE-2000 PFS; Nikon) equipped with a spinning-disk confocal unit (CSU10; Yokogawa) with 200 mW, 488 nm and 561 nm solid-state lasers (LMM5; Spectral Applied Research), electronic shutters, a cooled charge-coupled device camera (Cool-SNAP HQ2; Photometrics), and controlled by NIS Elements software (Nikon). Microtubule growth rates were determined by computational tracking of EB1-EGFP comets in time-lapse sequences. Comet detection was improved by reducing low frequency background variation with a homomorphic filter in the frequency domain (MetaMorph; Molecular Devices). Only tracks existing for at least 5 frames (2.5 sec) were used for analysis to bias against microtubules that did grow at an angle relative to the plane of focus. A 60 × NA 1.45 oil immersion lens (Plan Apo TIRF; Nikon) was used to image cell extensions by DIC microscopy and for immunofluorescence. To measure protrusion dynamics in DIC images, we used the manual tracking function in NIS Elements (Nikon). A 40 × objective lens (Plan Fluor ADM/ELWD; Nikon), and the large image function in NIS Elements were used to collect 6 × 6 stitched phase contrast time-lapse sequences. In these high-resolution phase contrast images, matrix tension tracks were readily visible as dark striations and were counted at indicated time points. Statistical analysis was performed in Excel (Microsoft) using the Analyze-it plug-in for Excel (Analyze-it Software, Ltd). Box-and-whisker plots show the median, the 1st and 3rd quartile (box), with whiskers extending to the furthest observations within ± 1.5 times the interquartile ranges of the 1st or 3rd quartile. P values were calculated from two-sample equal variance, 2-tailed t-tests. Images for presentation in figures and movies were contrast-enhanced as described.

Click here for additional data file.

Click here for additional data file.

Click here for additional data file.

Click here for additional data file.

Click here for additional data file.