Real-time two- and three-dimensional imaging of monocyte motility and navigation on planar surfaces and in collagen matrices: roles of Rho

We recently found that macrophages from RhoA/RhoB double knockout mice had increased motility of the cell body, but severely impaired retraction of the tail and membrane extensions, whereas RhoA- or RhoB-deficient cells exhibited mild phenotypes. Here we extended this work and investigated the roles of Rho signaling in primary human blood monocytes migrating in chemotactic gradients and in various settings. Monocyte velocity, but not chemotactic navigation, was modestly dependent on Rho-ROCK-myosin II signaling on a 2D substrate or in a loose collagen type I matrix. Viewed by time-lapse epi-fluorescence microscopy, monocytes appeared to flutter rather than crawl, such that the 3D surface topology of individual cells was difficult to predict. Spinning disk confocal microscopy and 3D reconstruction revealed that cells move on planar surfaces and in a loose collagen matrix using prominent, curved planar protrusions, which are rapidly remodeled and reoriented, as well as resorbed. In a dense collagen type I matrix, there is insufficient space for this mode and cells adopt a highly Rho-dependent, lobular mode of motility. Thus, in addition to its role in tail retraction on 2D surfaces, Rho is critical for movement in confined spaces, but is largely redundant for motility and chemotaxis in loose matrices.

Scientific RepoRts | 6:25016 | DOI: 10.1038/srep25016 is a key regulator of cytoskeletal dynamics in cells and is implicated in at least two aspects of amoeboid-like migration, retraction of the trailing edge on 2D surfaces and squeezing of the nucleus (principal geometrically limiting factor 5 ) through narrow spaces 6 .
The roles of the Rho subfamily in cell migration have been widely studied 3,[6][7][8][9][10][11][12] , and it has become clear that the relative importance of Rho depends on cell type, mode of migration and the 2D or 3D microenvironment. In an elegant model, Petrie et al. 10 proposed that fibroblasts read the extracellular matrix elasticity and can switch between low RhoA-ROCK-dependent, lamellipodial (nonlinear elastic matrix) and high RhoA-ROCK-dependent, lobopodial (linear elastic matrix) modes of migration (see also Friedl et al. 13 ). Although RhoA is considered to be the predominant Rho subfamily member involved in cell motility, we found in a knockout mouse study 14 that RhoB compensates for loss of RhoA, such that macrophages lacking either RhoA or RhoB had mild phenotypes, whereas cells lacking both RhoA and RhoB exhibited profound phenotypes, characterized by highly branched morphologies during random migration (largely due to impaired lamellipodial retractions) and exceedingly long trailing ends during directed migration. Chemotaxis efficiency was not impaired in macrophages lacking RhoA and RhoB (as well as, by default, RhoC), and, surprisingly, the velocity of the cell body, as well as in vivo monocyte/macrophage recruitment, were increased 14 . In the current study, we extended this work and used well-documented inhibitors to elucidate the roles of Rho-ROCK-myosin II signaling in the motility and chemotaxis of human monocytes on 2D surfaces and in both loose and dense (fibrillar) collagen type I matrices. We used two approaches, contrast and/or epi-fluorescence microscopy, which allowed time-lapse imaging for long durations (> 10 h), and sequential spinning disk confocal microscopy, which allowed high-resolution 3D reconstruction of cells in motion, albeit in a shorter time span (< 30 min).

Expression of Rho subfamily GTPases and roles of ROCK-myosin II signaling in monocyte motil-
ity and chemotaxis on a 2D surface. Using CD14 + cells (monocytes) purified by cell sorting, we could detect mRNA for RhoA, RhoB and RhoC (Fig. 1A). Furthermore, we could detect RhoA, RhoB and RhoC protein using Western blot (Fig. 1B). Thus, unlike mouse macrophages, which only express RhoA and RhoB 14 , human monocytes express all three members of the Rho subfamily. We next set out to explore the effects of pharmacological inhibition of Rho, ROCK and nonmuscle myosin II (NMMII) on monocyte motility and chemotaxis, as indicated in Fig. 1C. Initially, we performed 2D chemotaxis experiments ( Fig. 2A), as previously described for mouse macrophages 15,16 . However, human monocytes migrated much more rapidly than macrophages (~4 μ m/min versus ~1 μ m/min), and, thus, we analyzed a shorter time window, 4 h rather than 6 h. Under control conditions, monocytes migrated robustly towards fMLP and mean cell velocity was unaffected in the presence of a ROCK inhibitor (Y-27632) (Fig. 2B,C), but modestly reduced by the nonmuscle myosin II inhibitor S-blebbistatin. Moreover, Y-27632 and S-blebbistatin had no significant effect on chemotaxis efficiency (Fig. 2C). As indicated in Fig. 2A, Y-27632 induced elongated trailing ends in monocytes migrating in a chemotactic gradient. To quantify the effect of Y-27632 on cell morphology, we measured the circularity index and aspect ratio, after fitting an ellipse, of spontaneously migrating cells, imaged by spinning disk confocal microscopy (Fig. 2D,E). Consistent with a more elongated morphology, Y-27632 treatment induced a significantly decreased circularity index, and significantly increased aspect ratio, measured using two-dimensional projections.
3D morphology of monocytes migrating on a 2D surface. To establish a cell-type specific chemotaxis assay which also allows fluorescence imaging, we labeled human monocytes with Alexa Fluor 488-conjugated anti-CD14 antibodies (Fig. 3A). Time-lapse epi-fluorescence imaging provided better visualization of monocytes than phase-contrast imaging, and fluorescent cells appeared to move towards fMLP like fluttering leaves (see supplementary video 1), rather than like crawling macrophages. In sequential snapshots, individual cells appeared to extend one or more broad lamellipodia (generally towards the source of chemoattractant) in one moment and in the next moment exhibited either a compact or comma-shaped morphology (Fig. 3A,B). We used time-lapse spinning disk confocal microscopy to allow rapid confocal sectioning and 3D reconstruction of migrating monocytes (Fig. 3B-D). 3D reconstructed images revealed that, instead of lamellipodia (sheet-like protrusions on a 2D substrate), monocytes generate curved (scalloped) planar protrusions in varying orientations in space (Fig. 3B-D). These 3D membrane dynamics can best be appreciated in videos rather than still images. For example, supplementary video 2 (3D time-lapse image sequence) shows the cells shown in Fig. 3B viewed from above. Supplementary video 3 serves as a (time-lapse 2D imaging) reference sequence for the cells shown in Fig. 3C, whereas supplementary video 4 shows a time-lapse 3D reconstruction of the cells. Classical fibroblast-like lamellipodial protrusions could be detected (not shown), but curved planar protrusions (and membrane undulations) in various orientations predominated. Lack of classical lamellipodia could be appreciated by viewing cells from below (Fig. 3D), which revealed that, typically, only the cell body and the lower edge of planar sheets made contact with the 2D substrate. Supplementary video 5 shows 3D views the same cells (in Fig. 3D) rotated 360° on the x-axis. The low contact area, combined with the reorientations of curved planar protrusions and membrane undulations, may explain the fluttering movement observed in epi-fluorescence time-lapse recordings.
Phototoxicity and photodegradation of blebbistatin. One of the major end targets of Rho-ROCK signaling is nonmuscle myosin II (Fig. 1C), which can be inhibited by blebbistatin 17 , a racemic mixture of R-(+ )and S-(−)-blebbistatin (RS-blebbistatin). In initial chemotaxis assays, we found that illumination of monocytes (labeled with Alexa Fluor 488-conjugated anti-CD14 antibodies) with blue light (excitation: 450-490 nm) in the presence of RS-blebbistatin induced cell death, consistent with earlier reports that blue light photoinactivates blebbistatin and generates cytotoxic intermediates 18,19 . Consistent with the data of Kolega 18 , we found that blue light (450-490 nm) illumination induced changes in the absorption spectrum of S-blebbistatin (Fig. 4A), solubilized in a 1:1 mixture (by volume) of DMSO and PBS (phosphate-buffered saline). We extended this work, and found that ionized S-blebbistatin, the active enantiomer, could not be detected by mass spectrometry after blue light illumination, indicating that blebbistatin phototoxicity is associated with photodegradation of the molecule (Fig. 4B).
More recently, the combination of blue light and blebbistatin has been reported to be toxic to human cancer cell lines, and the authors suggested that blebbistatin may be a potential anti-cancer agent 20 . To assess the phototoxicity of blebbistatin in primary human monocytes, we used the cell-impermeant nucleic acid stain Yo-Pro-1 as a probe to detect loss of cell membrane integrity during live-cell recordings (Fig. 4C,D). Using spinning disk confocal microscopy, cells were excessively illuminated with blue (488 nm) light by taking serial z-stacks (20 optical sections of 0.5 μ m thickness) at a rate of 4 stacks/min. In the presence of 50 μ M RS-blebbistatin, 488 nm illumination induced Yo-Pro-1 staining in more than 95% of cells during the 45 min recording period (Fig. 4C,D). Note that the images shown in Fig. 4C were obtained after moving the microscope stage partially to the right in order to reveal cells (seen on the left half of the field) which were not scanned with 488 nm light; these non-illuminated cells were not used in the data analysis. In the absence of RS-blebbistatin, 488 nm illumination alone was not sufficient to induce Yo-Pro-1 permeability during the 45 min recording period (Fig. 4D).
Monocyte motility and navigation in loose collagen type I matrices. Following the canonical leukocyte adhesion cascade, which involves selectin-mediated tethering and rolling, as well as the activation of integrins, monocytes migrate across the endothelium into the interstitium, a highly heterogeneous 3D environment 21 . Fibrillar collagen, especially type I, is the most abundant extracellular matrix component, and a wide range of lattice densities, from loose to dense, are found in the same tissue 21 , whereas collagen matrices reconstituted in vitro exhibit a narrow range. In this study, we tested two collagen type I concentrations, which were intended to represent loose (see Fig. 5A) and dense fibrillar collagen lattices. Monocytes were prestained with fluorescent (Alexa Fluor 488-conjugated) anti-CD14 antibodies (or Alexa Fluor 594-conjugated antibodies, in the case of experiments using blebbistatin) and imaged by time-lapse epi-fluorescence microscopy (Fig. 5A). In a loose collagen matrix (0.8 mg/ml collagen type I), control cells moved robustly towards fMLP ( Fig. 5B; supplementary video 6). Monocytes fortuitously in contact with the bottom of the chemotaxis slide were used to set the level of focus, and, thus, most of the time cells moving through the loose matrix appeared as fuzzy (out of focus) objects (supplementary video 6). In the presence of inhibitors of Rho (TAT-C3), ROCKs (Y-27632) or nonmuscle myosin II (blebbistatin), monocyte motility was modestly decreased, but there was no significant impairment in chemotactic navigation ( Fig. 5B-D). Interestingly, in parallel experiments, using spinning disk confocal microscopy, we observed that monocytes treated with TAT-C3 developed exceedingly elongated trailing ends when crawling along a 2D surface, but not within a loose 3D matrix (supplementary video 7).  3D morphology of monocytes migrating in a loose collagen type I matrix. As in the case of 2D migration, we performed spinning confocal microscopy to visualize the mode of migration. Monocytes moved through a loose collagen matrix using curved, planar protrusions in varying orientations (Fig. 6A), as observed for cells moving on a 2D surface. The 3D morphology of the cells shown in Fig. 6A, at t = 0 s, can be nicely seen in supplementary video 8, which shows the cells in different spatial orientations. The mode of migration was not altered in the face of ROCK inhibition (Fig. 6B,C). A time-lapse sequence corresponding to the cell shown in Fig. 6B can be seen in supplementary video 9, and supplementary videos 10, 11 and 12 provide multiple 3D views of the cell at t = 0 s, t = 240 s and t = 360 s, respectively. The time-lapse sequence (supplementary video 9) also Monocyte motility and navigation in dense collagen type I matrices. Superresolution structured illumination microscopy showed that polymerized 2.4 mg/ml collagen type I produces a lattice of fibers with small pore sizes (Fig. 7A). In a chemotactic fMLP gradient, fluorescently labeled monocytes robustly migrated through such dense matrices (Fig. 7A,B), whereas motility was profoundly decreased by inhibition of Rho-ROCK signaling (Fig. 7B) or nonmuscle myosin II (Fig. 7C). Surprisingly, the mean velocity of directed monocyte motility in a dense matrix was ~4 μ m/min (Fig. 7D), similar to mean values found on a 2D surface (Fig. 2C) or in a loose matrix (Fig. 5D). However, unlike the case for 2D surfaces and loose 3D matrices, inhibition of Rho-ROCK-myosin II signaling impaired chemotactic navigation (Fig. 7D). In relation to this finding, it should  be stressed that inhibitors of Rho, ROCK and nonmuscle myosin II frequently caused cells to become trapped in the matrix.
3D morphology of monocytes migrating in a dense collagen type I matrix. As in the case for monocyte motility in a loose matrix, we used spinning disk confocal microscopy to determine the morphology of cells moving in a dense matrix (Fig. 8). Under these conditions, monocytes moved in a lobular fashion, such that the front end consisted of blunt protrusions (Fig. 8A), presumably driven by hydrostatic pressure. A time-lapse recording of the cells shown in Fig. 8A can be seen in supplementary video 13, and 3D views of the cells at t = 360 s are provided in supplementary video 14. In the presence of Y-27632, monocytes became stuck in dense matrices, as indicated in Fig. 7B (chemotaxis assays) and Fig. 8B. Three-dimensional views of the cells shown in  Fig. 8B, followed by a time-lapse sequence, are shown in supplementary video 15. Thus, Rho-ROCK signaling is critical for motility in a dense matrix, but Rho-ROCK signaling per se does not determine the basic mode of motility, planar protrusions versus lobular protrusions. Indeed, independent of ROCK signaling, the circularity index (measure of roundness) is decreased, and the aspect ratio (measure of elongation) increased, when cells are migrating in a dense matrix, compared to a loose matrix (Fig. 8C).

Discussion
As dramatically formulated by Couzin-Frankel 22 , "inflammation is a driving force behind chronic diseases that will kill nearly all of us". Monocytes accumulate at sites of inflammation, as nicely documented by intravital imaging in mouse models 23,24 , but in addition to promoting tissue healing and pathogen clearance, these cells may aggravate disease. This "dark side" is particularly apparent in the pathogenesis of atherosclerosis 25 . Here, we established in vitro 2D and 3D chemotaxis assays, and used spinning disk confocal microscopy, to investigate the mechanisms of human monocyte motility, and we explored the roles of Rho-ROCK-myosin II signaling on cell shape control, velocity and chemotactic navigation.
Aside from a recent elegant study using a chemoattractant point source (micropipette) and fluorescently labeled chemoattractant proteins 26 , data obtained from real-time monocyte chemotaxis assays are very scarce, as far as we are aware. We used Ibidi 2D chemotaxis μ -slides for both 2D assays, as previously used for mouse macrophages 14,16 , and 3D assays, which necessitated addition of the chemoattractant before introduction of the cells. In contrast to resident mouse peritoneal macrophages, freshly isolated human monocytes did not readily flatten onto the ibiTreat (tissue culture treated) surface of μ -slides, and moved like fluttering leaves, with a mean velocity of ~4 μ m/min, towards chemoattractant. Under these low adhesion 2D migration conditions, inhibitors of ROCK and nonmuscle myosin II decreased monocyte velocity by ~40%, whereas chemotactic efficiency was unaffected. In macrophages completely lacking the Rho subfamily of GTPases, chemotactic navigation is similarly unaffected, whereas cell velocity is increased by ~50% 14 . Monocytes occasionally developed elongated trailing ends, but much less dramatically than pan-Rho knockout mouse macrophages 14 or macrophages treated with ROCK inhibitors 15 , presumably reflecting weak (monocytes) versus strong (macrophages) cell-substrate adhesions. Notably, in the presence of inhibitors of Rho-ROCK-myosin II, elongated trailing ends were not observed in cells migrating in loose or dense collagen type I matrices, except when cells transiently came in contact with the floor of the chemotaxis slide. This reinforces the notion that strong integrin-mediated adhesions are required for Rho-ROCK inhibitors to induce the development of elongated trailing ends in migrating cells.
In single frames of time-lapse epi-fluorescence recordings, human monocytes typically exhibited one of three basic morphologies (compact, fan-like or comma-like), and in time-lapse videos the cells appeared to flutter like leaves towards chemoattractant. Indeed, 3D reconstructions of z-stacks obtained by sequential spinning disk confocal microscopy indicated that the cells generated leaf blade-like protrusions tilted away from the x-y plane (corresponding to fan-like morphologies in x-y projection views) or rotated towards an orthogonal plane (corresponding to comma-shaped cells in x-y projection views). In other words, on uncoated 2D plastic surfaces, monocytes move using highly dynamic, three-dimensional protrusive structures, rather than crawling along the 2D substrate using lamellipodia. This raises the question whether monocytes can swim, as demonstrated for Dictyostelium amoebae and neutrophils in experiments using a point source of chemoattractant and isodense media 27 . In favor of this possibility, monocytes undergo rapid cell shape changes, and the pushing of curved planar protrusions against fluid, as well as membrane undulations, may support swimming.
We used superresolution structured illumination microscopy, as well as confocal reflection microscopy (not shown), to confirm that the collagen type I used in our studies formed fibrillar lattices. The loose collagen matrix exhibited more heterogeneous pore sizes, such that cells would be expected to encounter both low-to high-density fibrillar regions, as found in vivo 21,28 . In both loose and dense fibrillar collagen matrices, monocytes moved rapidly along fMLP gradients, as in the case of 2D assays, whereas inhibition of Rho-ROCK-myosin II signaling modestly decreased velocity in a loose matrix, but severely impaired motility in a dense matrix. As in the case of 2D motility, spinning disk confocal imaging revealed that monocytes moved in a loose matrix using curved planar protrusions. In contrast, in a dense matrix, virtually all cells moved in a sausage-like (or worm-like) fashion, involving blunt lobular protrusions. The high sensitivity to inhibition of Rho-ROCK-myosin II signaling suggests that monocytes switch from planar protrusion (pseudopodia) mediated motility to pressure (hydraulic) driven motility when the matrix pores becomes tight. Thus, human monocytes, like mouse dendritic cells 6 , use Rho-ROCK-myosin II signaling to "squeeze" through tight spaces. How a monocyte senses that it is in a tight matrix is unclear. One possibility is that membrane tension, the "sensor" for being in a tight matrix, is generated when the cell body becomes stuck. Such a sensor has recently been suggested to suppress lamellipodial membrane extensions on the trailing edge of migrating cells 29 . G 12 /G 13 -coupled receptors provide a potential link between membrane tension and Rho activity, although, at least on 2D surfaces, G 12 /G 13 double knockout macrophages do not develop elongated trailing ends (unpublished data) in contrast to pan-Rho-deficient cells 15 .
In conclusion, monocytes have been neglected in real-time imaging studies of motility and chemotaxis, although these cells play a pivotal role in human disease. We used 2D and 3D chemotaxis assays, as well sequential high-resolution 3D imaging, to better understand how human monocytes move and navigate, and to elucidate the roles of Rho-ROCK-myosin II signaling. In all conditions tested (2D substrate, 3D loose matrix and 3D dense matrix), monocytes moved with modestly high velocity and efficient chemotactic navigation, although the transition from loose to dense matrix caused the cells to switch from a fluttering leaf-like mode to a highly Rho-ROCK-myosin II-dependent lobular mode. Notably, the 3D membrane dynamics underlying the so-called fluttering leaf-like mode of motility could not be predicted from x-y projection views, which stresses the importance of sequential 3D (4D) imaging for understanding how cells move. It would be conceptually useful to investigate whether monocytes can swim in an isodense medium.

Materials and Methods
Materials. Recombinant TAT-C3 was prepared as previously described 30 , and used at a concentration of 50 μ g/ml. TAT-C3 is a cell permeable form of exoenzyme C3 transferase (from Clostridium botulinum), a Rhospecific inhibitor. It consists of C3 transferase fused to the protein transduction domain TAT (trans-activator of transcription), encoded by the human immunodeficiency virus gene Tat 31 . Y-27632 (Sigma-Aldrich, Germany), dissolved in DMSO to give a 30 mM stock solution, was used at 30 μ M. Racemic RS-blebbistatin, R-(+ )-blebbistatin and S-(− )-blebbistatin were obtained from Tocris Bioscience (Bristol, United Kingdom) and dissolved in anhydrous dimethyl sulfoxide (DMSO) to yield 50 mM stock solutions.

Ethics statement.
Methods were carried out in accordance with the approved guidelines of the University of Münster. All experimental protocols were approved by the local ethics committee of the University of Münster. Informed consent was obtained for the isolation of peripheral venous blood (15 ml) from donors. 2D chemotaxis assays. Chemotaxis assays were performed as previously described 14,32 , except that cells were allowed 2 h, rather than 3 h, to adhere to the narrow channel (observation area) connecting the two 40 μ l reservoirs of the uncoated (ibiTreat) μ -slide chemotaxis chamber (Ibidi, Martinsried, Germany). Chemotaxis slides were gently filled with medium by rotating the volume adjustment sleeve of Eppendorf Research Plus pipettes (2-20 μ l and 10-100 μ l). After filling, 15 μ l medium containing fMLP (formyl-Met-Leu-Phe) and Patent Blue V was drawn into one of the two 40 μ l reservoirs. After diffusion in the reservoir, the final concentration of fMLP was 20 nM. Patent Blue V (final concentration, 0.003%) served as a visual indictor of gradient formation.
The observation area was imaged by phase-contrast and/or fluorescence microscopy (excitation: 450-490 nm; emission: 500-550 nm) via a 10x/0.3 objective lens. In the case of experiments using blebbistatin, blue light illumination (450-490 nm) induced cell death. To avoid blebbistatin-mediated phototoxicity, we switched to Alexa Fluor 594-conjugated anti-CD14 antibody labeling and excited cells with green light (excitation: 532-558 nm; emission: 570-640 nm), which is not absorbed by the drug. Images were captured every 2 min for 12 h, and cell migration tracks between 3 and 7 h were analyzed with ImageJ (National Institutes of Health) using a manual tracking plugin and the chemotaxis and migration tool from Ibidi. Twenty-five randomly selected cells (monocytes) were manually tracked in each chemotaxis experiment. The y-forward migration (chemotaxis) index was used as a measure of chemotactic efficiency. The index is obtained by dividing the net displacement along the y-axis by the total distance (in the x-y plane) migrated by the cell (values range from − 1 to + 1).
Typically, 100 μ l cell suspension was mixed with 100 μ l of the 2x concentrated (and pH neutralized) collagen solution. Next, 8 μ l of the cell suspension was drawn into the narrow (1000 μ m × 2000 μ m × 70 μ m) channel of a prefilled and uncoated (ibiTreat) μ -slide chemotaxis chamber (Ibidi) using an Eppendorf Research Plus (2-20 μ l) pipette. Immediately before adding the cell suspension, chemoattractant (fMLP) was drawn into one of the two reservoirs.