Experiments were carried out according to the guidelines of the European Community Council Directives of January 1st 2013 (2010/63/EU) and of the local animal welfare committee, and all efforts were made to minimize the number of animals used and their suffering.

Primary cortical astrocyte cultures were prepared as previously described (Koulakoff et al., 2008). Briefly, brains were removed from CD1 newborn pups (P1-P3) and the cortices were dissected in cold PBS-glucose (33 mM). Meninges were carefully withdrawn and cortices were mechanically dissociated. Astrocytes were seeded on poly-ornithine coated glass coverslips in DMEM containing 10% fetal calf serum, 10 U/ml penicillin, and 10 μg/ml streptomycin (GIBCO) and incubated at 37°C, 5% CO₂. After 1 week, once cells have reached confluency, 1 μM of cytosine-arabinoside was added to the cell culture during 3 days to eliminate proliferating microglial cells. Medium was then changed every 3 days and cells were used after 2–3 weeks in culture. Confluent astrocytes were wounded by scraping monolayers with a 20 μl pipette tip (~300 μm in width) and imaged 6–8 h after the scratch (Etienne-Manneville, 2006).

STED imaging was performed using a home built STED microscope system with pulsed excitation and depletion, and time-gated detection described previously (Mahou et al., 2015). In brief the system was developed on a commercial point scanning microscope (RESOLFT, Abberior Instruments) based around a microscope frame (IX83, Olympus), a set of galvanometer mirrors (Quad scanner, Abberior Instruments) and a detection unit consisting of two avalanche photodiodes (SPCM-AQRH, Excelitas Technologies). Images were acquired with a 100X/1.4 NA oil immersion objective lens (UPLSAPO 100X, Olympus). STED excitation and depletion pulses are generated from the same titanium-sapphire oscillator (Mai Tai HP, Spectraphysics) operating at 765 nm. Laser power is divided between two optical paths. In the excitation path a supercontinuum is generated by pumping a photonic crystal fiber (SCG800, NKT photonics). The excitation wavelength is selected from this supercontinuum using a bandpass filter. Depletion pulses are taken directly from the titanium sapphire (λ = 765 nm). Both excitation and depletion pulses are dispersed to durations of 50–70 ps and 100–300 ps respectively through propagation in polarization maintaining single mode fibers. The arrival times are synchronized using an optical delay line. The depletion beam is spatially shaped into a vortex beam using a spatial light modulator (SLM; X10468-02, Hamamatsu). The Fluorescence emission was descanned. A custom dichroic mirror (DM; Abberior Instruments) and bandpass filter (ET685-70) were used. The depletion pattern is generated by displaying a helical phase mask on the SLM. This is also used to correct for aberrations particularly defocus and spherical aberrations induced by cell culture medium. Acquisition was controlled using the Imspector Image Acquisition software (Andreas Schönle, Abberior Instruments GmbH, Göttingen, Germany). STED excitation power was 10–20 μW and depletion power was 100–150 mW (measured at the back aperture). The pixel size used throughout was 50 nm (xy) and 500–750 nm (z, depending on cell thickness) cell thickness was between 2–3 μm. Maximum intensity projections are presented throughout. Time gates of 1.719 ns after the excitation pulse were used.

Confocal images of GFP transfected cells were acquired after STED imaging on the same set up. Fluorescence excitation is from a 488 nm laser source (Cobolt 06-MLD-488 nm, Cobolt). A DM (zt 594 RDC, Chroma) and emission filter (Brightline HC 550/88, Semrock) are used in detection.

AFM was performed on a commercial system (Bioscope RESOLVE, Bruker) and all measurements were performed using Nanoscope software (Bruker). PeakForce QNM-Live Cell probes (PFQNM-LC, Bruker AFM probes) were used throughout (Schillers et al., 2016). The deflection sensitivity of the probe was measured at the start of the experiment in situ, the spring constant of the probe was pre-calibrated and the no-touch calibration option in nanoscope software was used. Cells were imaged in the culture medium containing 10% HEPES. The sample stage was heated at 37°C. The AFM imaging mode was “Scanasyst in fluid” from the Nanoscope software—live cell and peak force set points are chosen between 200–600 pN for optimal force measurements. Typical values are 300 pN.

In this work a sample scanning AFM is used and fluorescence images are obtained from scanning the STED beam over the sample using galvanometric mirrors. The MIROview software (Nanoscope, Bruker) is used for correlative imaging. First the pixel sizes for optical imaging and AFM imaging are calibrated. The AFM translates the sample between three known xy positions and a fluorescent bead sample is imaged in each position. A common feature in each image is selected by the user. During imaging the AFM tip is manually centered on the optical field of view by translating the AFM stage. An optical image of the sample is then acquired. This image is imported into the MIROview software and AFM fields of view are selected on the optical image. Features which appear in both imaging modalities (e.g., borders of cells) are used as fiduciary markers and the optical image is translated, in software, such that both images are overlaid.

STED and confocal images are deconvolved using the Richardson-Lucy algorithm (deconvolution lab program in FIJI). The STED point spread function (PSF) is an average of PSFs measured using 20 nm beads. The confocal PSF is simulated using the package “PSF generator” available for FIJI.

Second order flattening is applied to AFM topography images to remove tilt and bow.

Stiffness values quoted are an estimate of the cone-sphere modulus taking into account the AFM tip size, shape and deflection sensitivity. Average values are measured for an 18 μm field of view (256 × 256). Where the field of view includes two cells then separate averages are taken for each cell. Values quoted are the average of all stiffness measurements in that cell. Fibritool analysis is used to measure the directionality of actin and tubulin filaments in images as an angle between −90° and 90°.

All data are expressed as mean ± SEM obtained from at least three independent experiments. n refers to the number of cells. Statistical significance for between-group comparisons was determined by unpaired or paired t-tests. Mann Whitney non parametric test has been used for discrete variables. Statistical analysis was performed in GraphPad InStat.

The STED microscope was developed for live cell imaging (Figure 1A). In particular, a pulsed laser at 765 nm was chosen for depletion, resulting in reduced phototoxicity as cells have low autofluorescence at this wavelength. To compensate for spherical aberrations induced by the change in refractive index between immersion oil and cell culture medium, a SLM was used to precompensate for wavefront aberrations. This system has been previously reported and achieves resolutions of up to 50 nm (Mahou et al., 2015; Fusco et al., 2016). Live cells were labeled with SiR-actin or SiR-tubulin (Lukinavičius et al., 2014). These organic dye labels have been previously reported as STED dyes and feature low cytotoxicity and high photostability for imaging intact cells. STED images allow the resolution of individual actin or tubulin filaments, compared to confocal microscopy (Figure 1B).

For live cell AFM imaging (Figures 1C,D), probes were chosen for imaging cells with varying heights. PeakForce QNM-Live cell probes (PFQNM-LC, Bruker AFM probes) were used due to their 17 μm long tips which allows imaging of structures with varying heights without the cantilever coming into contact with the sample (Schillers et al., 2016).

AFM measurements use the PeakForce QNM mode. This mode minimizes the forces applied by the cantilever to the sample and acquires a force curve for each pixel (Figure 1A). A map of both height and stiffness (as the Young’s modulus) can be simultaneously measured from this imaging mode (Figures 1C,D). The AFM probe sensitive to structures below the cell membrane reveals a polarized structure in the height image. Stiffness measurements are measured simultaneously with height images and therefore directly correlate.

For correlative imaging the cantilever is mounted and the AFM scan head is translated such that it is centered on the STED field of view. STED images of the sample are acquired prior to AFM images. In our experimental set-up, height images can be correlated with optical images of the cytoskeleton for an 80 × 80 μm field of view. This allows a direct correlation of cytoskeletal structure with both topography and membrane stiffness for living astrocytes.

We first simultaneously assessed astrocyte topography and membrane physical properties using AFM. AFM height images revealed highly polarized structural elements within in vitro astrocytes (Figure 2A). In addition, probing astrocyte surface with AFM generated maps of stiffness throughout individual astrocytes, and revealed a mean cell stiffness of 27.9 ± 2.6 kPa (n = 38, Figure 2B). Notably, a wide range of cell stiffnesses was observed, likely reflecting astrocyte heterogeneity. Cell labeling with GFP transfection enabled clear identification of individual astrocytes, and did not alter apparent cell stiffness (23.2 ± 3.3 kPa, n = 12, p > 0.05, Mann Whitney test, Figure 2B).

To identify the nature of the topographic structures (Figure 2A), we simultaneously performed STED imaging of actin cytoskeleton using SiR-actin dye in live astrocytes. We found a similarly polarized distribution of actin fibers within confluent astrocytes (Figure 2A). In addition, actin was also highly organized in array of fibers throughout individual astrocytes. Actin fibers revealed by STED imaging strongly correlated with the astrocyte topology assessed with AFM (Figures 2A,C). In addition, actin fibers and the structural elements from the AFM display similar orientation, as assessed by Fibritool analysis (region 1: AFM: 67.2°, STED: 68.8°; region 2: AFM: 56.6°, STED: 57.4°, Figure 2A). To test the contribution of actin cytoskeleton to astrocyte membrane physical properties, we disrupted actin networks in astrocytes using cytochalasin D (30 μM, 1 h; Figure 2D). GFP labeling enabled delineating individual astrocytes and co-alignment of fluorescence and AFM images. We found that actin disruption decreases astrocyte stiffness (GFP: 23.2 ± 3.3 kPa, n = 12; GFP + cytochalasin D: 14.5 ± 1.9 kPa, n = 10, p < 0.05, unpaired t test, Figure 2E). Altogether these data show that actin cytoskeleton contributes to astrocyte topography and membrane physical properties.

To investigate the involvement of tubulin on astrocyte topography and physical membrane properties, we performed STED-AFM recordings, where tubulin was labeled using SiR-tubulin dye. The tubulin cytoskeleton displayed a different structure compared to actin fibers, with no clear polarized orientation within astrocytes (Figure 3A). Tubulin filaments imaged with STED did not correlate with astrocyte topology revealed by AFM (Figures 3A,B). In addition, tubulin cytoskeleton showed a different orientation compared to the structural elements from the AFM (region 1: AFM: 32.2°, STED: 51.3°; region 2: AFM: 29.0°, STED: 40.0°, Figure 3A). To test the involvement of tubulin elements in astrocyte stiffness, we depolymerized tubulin structures in astrocytes using nocodazole (16 μM, 1 h; Figure 3C). GFP labeling enabled identification of individual astrocytes despite loss of tubulin structure. In contrast to actin, tubulin elements had no effect on astrocyte stiffness (GFP: 23.2 ± 3.3 kPa, n = 12; GFP + nocodazole: 26.2 ± 5.5 kPa, n = 7, p > 0.05, Mann Whitney test, Figure 3D).

Cytoskeleton remodeling occurs in migrating astrocytes. To induce astrocyte polarized migration, we used the well-established scratch assay in cultured astrocytes (Etienne-Manneville, 2006). Such assay generates astrocytes protrusions migrating in a direction perpendicular to the scratch through a well-defined integrin signaling activating small G-proteins (Etienne-Manneville and Hall, 2001). We found that during scratch-induced migration, actin fibers concentrate in a polarized fashion at the leading edge of astrocyte migrating protrusions (Figure 4A). Noteworthy, this was associated with a reduced organization of actin at the basis of the protrusions (Figure 4A). On the contrary, tubulin showed increased organization characterized by polarization of filaments in the direction of the migration both at the basis and at the leading edge of the protrusion (Figure 4B). To decipher whether cytoskeleton remodeling induced by polarized migration alters at the subcellular level membrane properties of astrocyte protrusions, we performed membrane stiffness measurements using AFM. We found that stiffness was increased at the leading edge of astrocyte protrusion (30.5 ± 4.3, n = 16) compared to its basis (16.3 ± 2.9, n = 16, p < 0.001, paired t test, Figure 4C).