3D nanoendoscopy-AFM

Visualizing intra-cellular nanostructures of living cells by nanoendoscopy-AFM

1. Cell culture

HeLa cells (Cell Resource Center for Biomedical Research, Institute of Development, Aging and Cancer, Tohoku University) and BALB/3T3 mouse fibroblasts (JCRB Cell Bank, Japan) were cultured in Dulbecco's Modified Eagle's Medium (DMEM, Fujifilm Wako Pure Chemical Corporation), supplemented with 10% FBS (fetal bovine serum, Biosera) and 1% PS solution (penicillin-streptomycin, Fujifilm Wako Pure Chemical Corporation). The day before the cell penetration experiments, cells were harvested from a Petri dish with 0.05% trypsin/EDTA for 2 minutes at 37 °C and centrifuged at 1400 rpm for 3 minutes. Then, cells were seeded onto a 35 mm plastic cell culture dish (TPP, Techno Plastic Products AG, Trasadingen, Switzerland), and cultured with DMEM for 24 hours. Prior to the measurements, the cell culture medium was replaced with CO₂-independent Leibovitz L-15 medium (Fujifilm Wako Pure Chemical Corporation), supplemented with 5% penicillin-streptomycin solution. The plastic Petri dish with the cultured cells was placed inside the Petri dish heater (JPK) preheated to 37 °C.
 

2. AFM needle-like nanoprobe fabrication

The long AFM microcantilever (500 µm length, 0.3 N/m spring constant) of an OPUS 3XC-GG cantilever chip (MikroMasch) and an ATEC-ContAu cantilever (450 µm length, 0.2 N/m spring constant, Nanosensors) were used to microfabricate the nanoprobes. Their silicon tetrahedral tip was milled by the focused ion beam (FIB) technique in a Helios G4 CX Dual Beam system (FEI, Thermo Fisher Scientific, USA). The cantilever tip was sequentially milled alternating its front and lateral sides, reducing gradually its diameter until the desired value. The accelerating voltage of FIB was kept at 16 KV at a working distance of 13 mm, reducing gradually the current from 1.3 nA to 15 pA when the nanoneedle’s diameter reduces to avoid tip damaging and increase milling resolution. The nanoneedles were FIB micro-fabricated with an angle of 10 degrees between their long axis and the cantilever plane, to correct the mounting angle on the AFM cantilever holder. 

The amorphous carbon (a-C) nanoneedles fabrication on Olympus BL-AC40TS cantilevers was performed by the electron beam deposition (EBD) technique, in a Helios G4 CX Dual Beam system from FEI (Thermo Fisher Scientific, USA). Naphthalene (C₁₀H₈) was used as gas precursor, injected onto the substrate surface through a nearby gas-injection system (GIS), and dissociated by the electron beam irradiation, producing a a-C deposition film on the same area of the scanned beam. The focused electron beam current was set to 0.17 nA at 15 kV acceleration voltage, with a working distance of 4 mm. The length of the nanoprobe is controlled by means of the time deposition, which was previously calibrated (in our system, 6 minutes of deposition corresponds to a length of 3.5 µm). The cantilever tip was truncated with the FIB before the nanoprobe fabrication. Then, the nanoneedle was deposited on the surface of the truncated tip. To increase the adhesion of the nanoneedle to the truncated cantilever tip, a few nanometers of a-C was deposited at the surface before the nanoprobe microfabrication. The nanoneedles were also growth with an angle of 10 degrees to correct the mounting angle on the holder. 

We have been extremely careful after every FIB milling or EBD deposition, minimizing the transferring time from the SEM/FIB system to the AFM cantilever holder immersed in liquids, reducing thus the deposition of contaminants on the nanoneedle surface. However, to further reduce the adhesion of contaminants to the nanoprobe surface, tailored coating would be required, such as anti-fouling molecular coatings, which may decrease cell damage during tip insertion and reduce nanoprobe contamination with biomaterials coming from the cytosol.
 

3. 3D nanoendoscopy-AFM

Cell penetration measurements were performed in a JPK Nanowizard IV BioAFM (Bruker Nano GmbH, Berlin, Germany). The nanoneedles were inserted into living cells at a speed raging from 10 to 50 µm/s, recording the cantilever’s deflection during the process, until a specific set-point is reached, retracting completely then the nanoprobe at the same speed. The 3D nanoendoscopy-AFM experiments were carried out in two different ways: 1) the nanoneedle above the cell is vertically moved down until a specific vertical force set point is reached (5-10 nN), retracting afterwards (Qi-mode); 2) similarly to the previous mode, the nanoneedle is vertically moved down, but in this case the cantilever is simultaneously oscillated with an oscillation amplitude of around 3 nm at its second resonance frequency (35 kHz ~ 50 kHz), until the cantilever’s oscillation amplitude is decreased by 80%, retracting the nanoprobe after (AC-mode force maps). In our experience, oscillating the cantilever while measuring helps to overcome the adhesion of the fibers to the nanoprobe. Indeed, using a 3D suspended fiber model, we confirmed that the friction force is greatly reduced when the second resonance vibration is excited. As the cantilevers used in the experiments presented low resonance frequencies (around 6 kHz) due to their low spring constant, we have used the second resonance mode to avoid the high noise level and the spurious resonances around the first resonance mode, produced by the cantilever’s acoustic excitation All the AFM experiments were performed at 37 °C. Sensitivity and cantilever stiffness were calibrated from its thermal noise spectrum. The 3D nanoendoscopy-AFM maps were processed with a custom built LabView software and visualized with Vortex software.

The HeLa cell displayed in Fig. 1E was measured with a FIB milled OPUS 3XC-GG cantilever, presenting a spring constant of 0.38 N/m. The force set point was 10 nN and the measured volume 40 × 40 × 8 µm³ with 64 × 64 × 1200 pixels. The vertical speed used was 50 µm/s. The HeLa cell depicted in Fig. 1F was imaged with a FIB milled ATEC-ContAu cantilever, with a spring constant of 0.2 N/m. The force set point was 5 nN and the measured volume 10 × 10 × 6 µm³ with 32 × 32 × 3000 pixels. The vertical speed used was 20 µm/s.

The images of BALB/3T3 mouse fibroblasts shown in Fig. 3 were acquired with a FIB milled OPUS 3XC-GG cantilever, presenting a spring constant of 0.29 N/m. The excitation frequency was set at 38.23 kHz presenting a free oscillation amplitude of around 6 nm. The amplitude set point was fixed at 4.5 nm. The measured volume was 4.5 × 4.5 × 3.5 µm³ with 90 × 90 × 3500 pixels. The vertical speed used was 30 µm/s.
 

4. 2D nanoendoscopy-AFM

2D imaging inside cells were also carried out in a JPK Nanowizard IV BioAFM, using the AC tapping mode. In this case, the cantilevers were oscillated close to their resonance frequency. Once the cantilever tip is placed above the cell area to be measured and before imaging the inner cell membrane, a force distance curve is performed, to exactly know where the upper and the bottom cell membrane are placed. After, the nanoneedle is introduced inside the cell at around 100 nm above the bottom cell membrane by manually setting the Z position of the piezo, waiting for 2 min to stabilize the nanoprobe inside the cell. Then, the feedback is set on at an oscillation set point, so the tip is tapping the bottom membrane. Finally, when the imaging is over, another force distance curve is performed to control if the cell structure suffered a fatal damage, comparing the height of the cell before and after the 2D imaging. Sensitivity and cantilever stiffness were calibrated from its thermal noise spectrum.

The images of the BALB/3T3 mouse fibroblast displayed in Fig. 4 were imaged with an Olympus BL-AC40TS cantilever with an a-C nanoneedle growth with the EBD technique, presenting a spring constant of 0.1 N/m. The excitation frequency was set at 34.4 kHz presenting a free oscillation amplitude of around 20 nm. The amplitude set point was fixed at 10 nm. The measured areas were 1.0 × 1.0 µm² (Fig. 4F) and 0.5 × 0.5 µm² (Fig. 4G) with 256 × 256 pixels. The scan rates were 1 Hz for Fig. 4F and 2 Hz for Fig. 4G.

Cantilever thermal vibration spectra measured outside and inside the cell (Fig. S10) show that the changes in the Q factor (2.4 to 1.96) and the resonance frequency (6.25 kHz to 6.20 kHz) were not significant, allowing the stable operation of 2D nanoendoscopy-AFM in dynamic mode. 
 

5. Cell viability assay

Cellular viability was assessed using Cellstain Double Staining Kit (Dojindo) which is a combination of two fluorochromes, calcein-AM and propidium iodide (PI). HeLa cells were incubated in Leibovitz's L-15 Medium (no phenol red, Thermo Fisher Scientific) including 0.6 µM Calcein-AM and 1.5 µM PI for 15 min at 37 °C before AFM imaging. We changed the medium to the Leibovitz's L-15 Medium (no phenol red) supplemented with 1% penicillin/streptomycin for AFM imaging. After AFM imaging, fluorescence images were acquired every 10 minutes for 250 minutes for 3D nanoendoscopy-AFM, and for 300 minutes for 2D nanoendoscopy-AFM. Finally, H₂O₂ was added to the medium (final concentration 0.1%) as control to kill the cells, acquiring images every 10 min. We used an inverted fluorescence microscope (Nikon Eclipse Ti2) with an EM-CCD (ANDOR iXon Ultra 888) and 20x/0.65NA lens were used on the assay. Calcein-AM was excited at 488 nm and detected with a 525 ± 25 nm bandpass filter. PI was excited at 540 nm and detected with a 610 ± 25 nm bandpass filter. Cell viability values were calculated dividing Calcein and PI intensity signals, plotting the ratios over the time.

3D nanoendoscopy-AFM cell viability results displayed in Fig. 2 and Video S2 were measured with a FIB milled OPUS 3XC-GG cantilever, presenting a spring constant of 0.38 N/m. The parameters of the 4 experiments are shown below:

1. Force set point = 10 nN; volume = 2 × 2 × 7 µm³; 64 × 64 × 3000 pixels; vertical speed = 20 µm/s.
2. Force set point = 10 nN; volume = 2 × 2 × 10 µm³; 64 × 64 × 3000 pixels; vertical speed = 20 µm/s.
3. Force set point = 10 nN; volume = 40 × 40 × 8 µm³; 64 × 64 × 1200 pixels; vertical speed = 50 µm/s.
4. Force set point = 10 nN; volume = 2 × 2 × 7 µm³; 64 × 64 × 3000 pixels; vertical speed = 50 µm/s.

2D nanoendoscopy-AFM cell viability results depicted in Video S4 of the Supplementary Materials were measured with the parameters presented below:

1. Set point = 6.3 nm; frequency = 30.7 Hz; area = 0.2 × 0.2 µm²; 128 × 128 pixels; scan rate = 1 Hz; 4 consecutive scans.
2. Set point = 5.7 nm; frequency = 30.7 Hz; area = 0.5 × 0.5 µm²; 128 × 128 pixels; scan rate = 1 Hz; 4 consecutive scans.
3. Set point = 5.3 nm; frequency = 30.7 Hz; area = 1 × 1 µm²; 128 × 128 pixels; scan rate = 0.5 Hz; 4 consecutive scans.
4. Set point = 5.5 nm; frequency = 30.7 Hz; area = 2 × 2 µm²; 128 × 128 pixels; scan rate = 0.5 Hz; 4 consecutive scans.
5. Set point = 5.8 nm; frequency = 30.7 Hz; area = 3 × 3 µm²; 128 × 128 pixels; scan rate = 0.5 Hz; 2 consecutive scans.
    

6. Fluorescence imaging of actin fibers

BALB/3T3 mouse fibroblasts were cultured on a µ-Dish 35 mm with a grid (Ibidi), and incubated in the Dulbecco's modified Eagle's medium (Gibco) supplemented with 10% fetal bovine serum (Biosera), 1% penicillin/streptomycin (Wako), and 1 µM SiR-actin (Cytoskeleton) for 1 hour at 37 °C before observation. After incubation, the medium is replaced by the Leibovitz's L-15 medium (no phenol red, Thermo Fisher Scientific) supplemented with 1% penicillin/streptomycin for AFM imaging. After AFM imaging, cells were immediately fixed in 4% paraformaldehyde/PBS solution (Wako). Fluorescence images were acquired using the laser scanning confocal microscope (Nikon Eclipse Ti2). We used a 100x/1.3NA lens with 50 nm z intervals. Excitation and detection were 640 and 685 ± 35 nm, respectively.

7. Process diagram flow of a 2D nanoendoscopy-AFM measurement

Fig. 1. Process diagram flow of a 2D nanoendoscopy-AFM measurement. The cantilever is oscillated close to its resonance frequency and placed above the cell area to be scanned. (1) Before imaging, a F-z curve is registered, to know precisely where the top and bottom cell membrane are located, recording the vertical deflection and the cantilever oscillation amplitude. The amplitude set point for the tip-sample distance regulation must be low enough to ensure the tip is tapping the cell bottom surface, but not too small to avoid applying large forces that may damage biological cell structures. In this case we chose a set point of 10 nm. The cell height is estimated around 600 nm. After, the nanoneedle is introduced inside the cell at around 100 nm above the bottom cell membrane by manually setting the Z position of the piezo with the feedback off, waiting for around 2 min to stabilize the nanoprobe inside the cell. (2) Then, the internal side of the cell membrane can be scanned for long periods. It is important to check if the absolute piezo height position on the first scanned image corresponds to the Z position at the set point on the F-z curve measured before, to be sure that we are measuring the bottom cell membrane. In this case, the values are similar: 3.6 µm on the F-z curve and around 3.7 µm on the 2D nanoendoscopy-AFM image, which confirms that we were measuring the internal side of the bottom cell membrane. (3) Finally, we repeat after finishing the AFM experiment again a F-z curve to verify it the cell height is the same than before the 2D nanoendoscopy-AFM measurements. As shown in this case, the cell height is still around 600 nm after measurements, a sign that the cell structure did not suffer large changes. Also, we need to check if the absolute piezo height position on the last scanned image corresponds to the Z position at the set point on the measured F-z curve, which leads again to similar values: 3.6 µm on the F-z curve and around 3.7 µm on the AFM image, reinforcing the fact we were measuring the bottom cell membrane.
 

8. Image processing protocol

Fig. 2. Colormap settings for cell 3D-maps. Colormaps and transparency settings for the 3D view of the cell membrane and the cytoplasmatic components and nucleus shown in Figs. 1G-H in the main text.

Prior to image visualization, we performed a pre-processing of the 3D data (background subtraction, smoothing, interpolation) in Python, LabVIEW (National Instruments) and Igor software (Wave Metrics) with custom-built programs, importing after the resulting 3D data into Voxler 3 (Golden Software) for visualization purposes. First, we import the raw AFM 3D data maps in Python. For images in Fig. 3, we have applied a Savitzky–Golay filter with a window of 13 points and order 3, which smooths the curves and facilitates afterwards the background subtraction in LabVIEW. During background subtraction, a smoothing of the signal is also performed, to remove unwanted crosstalk of the cantilever oscillation on the vertical deflection signal. An overall picture of the image processing followed is displayed in Fig. S14.

The 3D views of the 3D nanoendoscopy image shown in Figs. 1G-H were prepared using the colormaps as shown in Fig. S11. The pixel and physical sizes of the raw data were 64 × 64 × 1046 pix³ and 40 × 40 × 8.27 µm³, respectively. As the individual force curves constituting the 3D data have different lengths, the data includes NaN (not-a-number). 

1. Linear background subtraction: we fitted a linear function to the long-range part of the individual force curves and subtracted it.

(Cell membrane)

2. Visualizing the 3D force map from 0.2 – 1.4 nN with a transparency gradient as shown in Fig. S11B.

(Intra-cellular components)

3. Visualizing the 3D force map from 0.8 – 9.0 nN with a transparency gradient as shown in Fig. S11D.

(Substrate)

4. Visualizing the XY cross section taken from 3D force map from -1.0 – 9.0 nN without a transparency gradient as shown in Fig. S11F.

Figure 1G was prepared by overlaying these three maps for the cell membrane, intra-cellular components, and substrate. Meanwhile, Fig. 1H was prepared by changing the view angle of Fig. 1G and setting a clipping plane across the nucleus to show inside of the cell and nucleus.

Fig. 3. Colormap settings for cell volume 3D-maps. Colormap and transparency settings for the 3D view of the cell and the substrate shown in Fig. 1I in the main text.

The 3D views of the 3D nanoendoscopy image displayed in Fig. 1I were prepared as follows. The pixel and physical sizes of the raw data were 32 × 32 × 2560 pix³ and 10 × 10 × 5.55 µm³, respectively. As the individual force curves constituting the 3D data have different lengths, the data includes NaN (not-a-number). We applied 2D Gaussian Filter (size: 3 × 3, σ=0.6) to each XY cross section of 3D force image. 

1. Linear background subtraction: we fitted a linear function to the long-range part of the individual force curves and subtracted it.

2. Smoothing: the 3D force map was smoothed by filtering each XY cross section using 2D Gaussian filter (size: 3 × 3, s=0.6).

3. Interpolation: the 3D force map was interpolated by 1D linear interpolation in X and Y directions. The pixel sizes after interpolation are 64 × 64 × 2560 pix³.

4. Visualization: we have visualized 3D force map from 0.2 – 1.2 nN for the cell, and from 1.8 – 2.0 nN for the substrate with a transparency gradient as shown in Fig. S12B.

Fig. 4. Colormap settings for actin fibers 3D-maps. Colormaps and transparency settings for the 3D view of the upper and lower cell membranes and actin fibers shown in Figs. 3C-D in the main text.

Fig. 5. F-z curves processing diagram flow. Example of the force curves processed during the preparation of the 3D view of the upper and lower cell membranes and actin fibers shown in Figs. 3C-D in the main text.

The 3D views of the 3D nanoendoscopy image depicted in Figs. 3C-D were prepared as follows. The pixel and physical sizes of the raw data were 90 × 90 × 3736 pix³ and 4.5 × 4.5 × 4.24 µm³, respectively. As the individual force curves constituting the 3D data has different length, the data includes NaN (not-a-number). 

(Upper membrane)

1-1. Linear background subtraction: we fitted a linear function to the long-range part of the individual force curves and subtracted it.

1-2. Data cut: we removed the upper part of the 3D data, where most of the points are NaN. The pixel and physical sizes of the extracted image were 90 × 90 × 1868 pix³ and 4.5 × 4.5 × 2.12 µm³, respectively.

1-3. Decimation: Z pixel size was reduced from 1868 pix to 266 pix by averaging every 7 points.

1-4. Visualization: we visualized the force distribution from 68 – 113 pN with blue color scale to highlight the upper cell membrane with a transparency gradient as indicated in Fig. S13A-B.

(Lower membrane)

2-1. Smoothing: the decimated 3D force map, obtained in the above 1-3 step, was smoothed by averaging adjacent three points.

2-2. Differentiation: the force map was differentiated with respect to z.

2-3. Data Cut: the bottom 10 points are extracted from individual force curves.

2-4. Visualization: we visualized the force gradient distribution over -250 mN/m with white color scale to highlight the lower cell membrane with a transparency gradient as shown in Fig. S13C-D. Note that the data values in Fig. S13D should be divided by 8 nm, which corresponds to the Z spacing of the pixels.

(Actin fibers)

3-1. Exponential background subtraction: We fitted the following exponential function to the force curves in the decimated 3D force map (obtained in the above 1-3 step) and subtracted it.

When we performed the fitting, A and z₀ were fixed so that the curve goes through the point closest to the substrate in the force curve. In addition, the force values were used as a standard error of the fitting for giving force-dependent weighting. In this way, the lower force points have a higher weight in the fitting. These constraints and weighting are effective to suppress possible artifacts introduced by the background subtraction.

3-2. Remove low values: values less than 120 pN was replaced with NaN.

3-3. Smoothing: the force curves were smoothed by averaging seven adjacent points.

3-4. Differentiation: the force curves were differentiated with respect to z.

3-5. Visualization: we have visualized force gradient distribution over -400 mN/m with a transparency gradient as shown in Fig. S13E-F. Note that the data values in Fig. S13F should be divided by 8 nm, which corresponds to the Z spacing of the pixels.