Actin-based protrusions of migrating neutrophils are intrinsically lamellar and facilitate direction changes.

Cell lines

Polymerized cortical actin filament were labeled with a high-affinity actin binding domain from utrophin fused to the red fluorescent protein mCherry ( Belin et al., 2014 ), and plasma membrane labeled by fusing a palmitoylation sequence from Lyn kinase ( Inoue et al., 2005 ) to the green fluorescent protein mEmerald. Utrophin-mCherry HL-60 cells were derived by lentiviral transduction of cell line #CCL-240 obtained from the American Type Culture Center (ATCC), where the cell line’s identity was confirmed by STR. HL-60 cell lines were grown in medium RPMI 1640 supplemented with 15% FBS, 25 mM Hepes, and 2.0 g/L NaHCO3, and at 37C with 5% CO2. HL-60 cell lines tested negative for mycoplasma by both PCR and DNA staining. Lentivirus was produced in HEK293T grown in 6-well plates and transfected with equal amounts of the lentiviral backbone vector (a protein expression vector derived from pHRSIN-CSGW ( Demaison et al., 2002 ), by cloning the actin-binding domain of utrophin ( Burkel et al., 2007 ) followed by a flexible linker (amino acid sequence: GDLELSRILTR) to the N-terminus of mCherry), pCMV∆8.91 (encoding essential packaging genes) and pMD2.G (encoding VSV-G gene to pseudotype virus). After 48 hr, the supernatant from each well was removed, centrifuged at 14,000 g for 5 min to remove debris and then incubated with ~1×10 ∧ 6 HL-60 cells suspended in 1 mL complete RPMI for 5–12 hr. Fresh medium was then added and the cells were recovered for 3 days to allow for target protein expression, and expressing cells were selected by fluorescence-activated cell sorting (FACS). Membrane-mEmerald and Lifeact-mCherry cell lines were derived as above and fusing the palmitoylation sequence from Lyn kinase ( Inoue et al., 2005 ) to the green fluorescent protein mEmerald, and the Lifeact peptide ( Riedl et al., 2008 ) to mCherry, respectively. Prior to imaging, HL-60 cells were differentiated by treatment with 1.3% DMSO for 5 days. For two-dimensional migration, differentiated HL-60 cells were allowed to adhere to fibronectin-coated coverslips for 30 min before coverslip was moved to the imaging chamber. For three-dimensional migration, cells were overlayed onto coverslips containing pre-formed 1.7% collagen matrix polymerized from a 1:1 mixture of of unlabeled (Advanced Biomatrix catalog no. 5005) and FITC-conjugated (Sigma catalog no. C4361) bovine skin collagen, using standard protocols ( Sixt and Lämmermann, 2011 ). Cells were allowed to migrate into the network for one hour before removing the coverslip to the imaging chamber. Cells were imaged in 1 × HBSS supplemented with 3% FBS, 1 × pen/strep, and 40 nM of the tripeptide formyl-MLP (to stimulation migration). Treated cells were exposed to 10 µm CK-666 (Sigma) or DMSO carrier alone (control) for ten minutes prior to imaging.

Lattice light sheet microscopy of living cells

Time-lapse sequences generated by the lattice light sheet microscope comprised two-color image stacks, collected through entire cell volumes at 1.3 s intervals with minimal photobleaching and no evidence of phototoxicity. The spatial resolution of the resulting data was approximately 230 nm in XY and 370 nm in Z ( Chen et al., 2014 ). The imaging of HL-60 cells was carried out in lattice light sheet microscopy using Bessel beams arranged in a square lattice configuration in dithered mode, as in ( Chen et al., 2014 ). For 2D migration, differentiated HL-60 cells were allowed to adhere to fibronectin-coated coverslips for 30 min before coverslip was moved to the imaging chamber at 37°C. The data was acquired on a Hamamatsu ORCA-Flash 4.0 sCMOS camera, where the moving cell was imaged by exciting each plane with a 488 nm laser at ~10 µW (at the back aperture of the excitation objective) for 5 ms, with an excitation inner/outer numerical aperture of 0.55/0.48 respectively and a corresponding light sheet length of 15 µm. At each time point, the cells were imaged by sample scanning mode and the dithered light sheet at 400 nm step size, thereby capturing a volume of ~70 µm x 90 µm x 40 µm (corresponding to 672 × 908 × 201 pixels in deskewed data) every 1.2 s (which includes 1 s acquisition time and 0.2 s pause between each time point). There are 143 time points for the continuous imaging periods of 2.86 min in duration. For three-dimensional migration, mCherry- utrophin HL-60 cells were overlayed onto coverslips containing pre-formed 1.7% collagen matrix polymerized as described (14) using 50% FITC-collagen (Sigma) and allowed to migrate into the network for one hour before removing the coverslip to the imaging chamber. The cell moving inside the collagen was imaged by exciting each plane with a 488 nm laser (collagen) and 568 nm laser (HL-60) at ~20 µW and 10 µW (at the back aperture of the excitation objective) for 5 ms separately before moving to the next z plane, with an excitation inner/outer numerical aperture of 0.4/0.325 respectively and a corresponding light sheet length of 20 µm. At each time point, the cells were imagined by objective scanning mode and the dithered light sheet at 250 nm step size, thereby capturing a volume of ~70 µm x 36 µm x 35 µm (corresponding to 672 × 352 × 141 pixels in raw data) every 1.4 s (which includes 1.3 s acquisition time and 0.1 s pause between each time point). There are 250 time points for the continuous imaging periods of 5.83 min in duration. Image alignment, compression and normalization With the datasets we collected the imaged volume was typically five times larger than the cell along each axis so that the cell did not crawl out of view. To enable real-time playback from any vantage point we used compression and thresholding of the image data to reduce its size by as much as a factor of 100 utilizing that the cell at any instant only occupied a small portion of the imaged box. To compensate for scope drift we aligned static features (extracellular collagen filaments) of each three-dimensional image to the preceding time maximizing cross-correlation to correct microscope jitter. We normalized the intensity levels by shifting the mean to 0 and scaling the intensity for each three-dimensional image so enclosed cell volume was constant at a standard intensity value (arbitrarily chosen equal to 100). This corrected variations in normalization introduced by microscope deconvolution software and also compensated for photobleaching which reduced intensity levels by approximately a factor of 2 over the imaged time periods. Rendering images The observed cell protrusions undergo continuous change in shape and can emerge from all sides of the cell. To communicate these features in two-dimensional static images we exported the surface mesh from UCSF Chimera as a Wavefront (.obj) file, and imported them into Cinema4D, a three-dimensional animation software package. We used the shading technique of ambient occlusion (recessed areas appear dark) and the Sketch and Toon Shader both with default settings to clearly visualize the three-dimensional dynamic behavior ( Figure 1—figure supplement 3 ). Real-time viewing of the data with ‘vseries’ toolkit We developed ‘vseries’ software within UCSF Chimera to analyze the large crawling cell data sets. The ‘vseries’ toolkit includes command line implementation and a Volume Series GUI. It enables users to view and manipulate an ordered sequence of volumetric datasets, and apply analysis commands to some or all of the volumetric maps at once. We used three-dimensional interactive visualization to see all aspects of the cell motion allowing any feature to be examined at any time in the cell motion. The new software visualization and analysis capabilities have been distributed as the ‘vseries’ command of the UCSF Chimera visualization program. To mark features of interest we also developed three-dimensional interactive surface visualizations for viewing in a web browser with the ability to hand place markers (small spheres) on desired cell features, and annotate them with text descriptions and choice of color. This software enables collaborative online annotation of the data.

Quantification of protrusion thickness and flatness

To measure thickness of the rendered three-dimensional cells, surface meshes were imported into Cinema4D and markers were created to measure the top-to-bottom distance along the edge of a pseudopodial sheet. Flatness was measured in UCSF Chimera by first placing markers along the edge of the sheet using the ‘Markers’ function ( http://www.rbvi.ucsf.edu/chimera/current/docs/ContributedSoftware/volumepathtracer/framevolpath.html ) and then running a Python script that calculates the best fit plane and measures the distances from the plane to the markers. We also measured the thin edge of lamellar sheets intersecting raw lattice light sheet images at 90° (567 ± 64 nm on glass and 689 ± 86 nm in collagen) at half maximum fluorescence intensity). We also measured phalloidin-stained actin imaged using confocal microscopy (505 ± 49 nm for a sheet and 533 ± 123 nm for a rosette petal) intersecting the confocal plane at 90°, measured at half maximum fluorescence intensity.

Quantification of cell movement and protrusion activity

To quantify the motions we measured cell centroid position as a function of time to obtain speed, direction, and changes of direction. This capacity is now included as the ‘vseries measure’ command. For each volume in the volume series, the command calculates the centroid (x,y,z) coordinates, the distance from the previous centroid (‘step’), the cumulative distance along the piecewise linear path from the first centroid, the surface-enclosed volume, and the surface area. The results are saved in a text file. The centroid is the center of mass of the density map based on map regions above the threshold; we set the threshold to our already established normalized level of 100.

Quantification of protrusion activity

To automate identification of the irregularly shaped protrusions, we took advantage of the fact that utrophin-based cortical actin labeling entered nascent protrusions slowly while membrane labeling was always present in the protrusions. We computed difference maps between the imaged membrane and actin channels, applied spatial smoothing using a Gaussian filter with a standard deviation of 0.5 µm to reduce noise. The resulting enclosed volumes were marked as individual protrusions if they were greater than 1 µm 3 , less than 15 µm from center, and outside cell center radius or larger than 15 µm 3 . For each volume in the volume series we calculated the total volume of all the protrusions, the distance of each protrusion from the cell center, the angle between the line connecting the cell centroid to protrusion centroid and the cell centroid path, and the number of protrusions. This calculation was implemented as a Python script ( Figure 6—source code 1 ).

Quantification of turning points

Changes of cell direction appeared correlated with emergence of multiple protrusions in different directions. To examine this correlation we used principal component analysis to find the primary plane of motion for graphing a cell centroid path in two dimensions. The first method we used for quantifying direction changes was to calculate the sliding time window ratio of the path length over the Euclidean distance between end-points using an interval of 10 time points. A ratio greater than 2.5 was considered turning. Our second approach was a Savitzky–Golay filter to calculate an estimate of the derivative. Time points were considered turnaround points if the derivative was zero and the cell had traveled at least 1 µm from the last turnaround point. Both of these methods were written in R Studio ( Figure 6—source code 2 ). Conventional microscopy RICM, TIRF, and spinning disk confocal microscopy were performed on a Nikon Ti-E inverted microscope equipped with a Spectral Diskovery and an Andor iXon Ultra EMCCD. RICM illumination (also called ‘interference reflection microscopy’) was produced with a 50/50 beamsplitter (Chroma 21000) in the dichroic cube and a 550 nm excitation filter (Thorlabs FB550-10) after a halogen epi lamp. Sufficient neutral density filters were placed after the halogen lamp and the aperture diaphragm was adjusted to maximize the contrast of the RICM images. RICM and TIRF images were interleaved less than 500 ms apart by rotating the RICM cube in and out of the path every other frame. All hardware was controlled using Micro-Manager software ( Edelstein et al., 2010 ). Image analysis of RICM, TIRF, and spinning disk confocal data was performed using ImageJ unless otherwise noted. Scanning electron microscopy Jurkat T lymphocytes were obtained from the American Type Culture Collection (ATCC) who confirms cell line identity by STR, and were grown in suspension in RPMI 1640 (Gibco) supplemented with 5% fetal bovine serum (Atlanta Biologicals Inc), 2 mg/mL glucose, 1 mM sodium pyruvate (Gibco) and 50 uM beta-mercaptoethanol (Gibco), and were maintained at densities