The development of new methods for imaging living cells with nanometric resolution is one of the most addresses topics in modern biology. Atomic force microscopy (AFM) provides an unpaired resolution in the out of plane z coordinate, - often around 1nm and even better -, and in the xy plane about 10nm and less. AFM was successfully applied to robust living cells  however, since this technique is based on the occurrence of a physical contact with the sample which can be too invasive for soft samples, application to fragile living cells such as neurons, can be problematic [2-5].
Here we propose a simple but effective approach enabling the investigation of the dynamics of neuronal growth cone (GC) with nanometric resolution, combining AFM with time lapse Differential Interference Contrast (DIC) imaging. We applied this novel method to investigate the morphology of the GCs at the neurite ends in embryonic stem (ES) cell derived neurons [6-7]. During early development of the central nervous system (CNS), there is an exuberant outgrowth of projections which later need to be refined to achieve precise connectivity . Exuberant and/or erroneous neuronal connections need to be pruned to achieve precise connectivity. GC pruning occurs through retraction, degeneration, or a combination of both . Analyzing GCs topography with AFM and movements with time lapse DIC imaging we verified a new pruning mechanism, not previously reported in literature, in which GC are sequentially dissolved, with an initial flattening, followed by the formation of hole and a final fragmentation on the whole structure.
Cell Culture and AFM Imaging
ES-derived neuronal precursors were obtained using the protocol described previously in Ban et al. (2007) . Cells were plated on coverslips and induced to differentiate for 24 hours. The structure of the differentiating living GCs was analyzed by AFM (Nanowizard II instrument; JPK, Berlin, Germany) in tapping mode at 0.6 scan line per second and with a maximum of 256 scan lines, using soft cantilevers (up to 0.03 N/m). Cells were plated at a density of 3x104 cells/cm2 in order to obtain isolated GCs to avoid overlapping structures and coverslips with living cells were mounted into a stage (BiocellTM, JPK) maintaining temperature at 37°C, covered with 1.5 ml of differentiation medium.
AFM Live Imaging
We investigated GC dynamics of living neurons derived from ES-cells 24 hours after differentiation.
GC cycles of protrusion and retraction were observed. An example of a growing GC is displayed in figure 1a-c. The external surfaces were measured with AFM imaging and GC edges (see red arrow of the fig. 1c) could advance by 1-3 microns within 20 minutes. The height of their central domain increased from 150 to more than 300 nm and subsequently decreased to 200 nm (height profile fig. 1d). The observed growth of imaged GC indicates that the physiological dynamics was poorly affected by AFM scans and that the cantilever tip did not damage the GC. In this way, we were able to follow more than 10 GCs for hours and all of them had compact and smooth surfaces.
Growth is be followed by retraction , indeed an example of a retracting GC is displayed in figure 1 e-g. Here AFM images were collected repetitively over 2-4 hours (160 min between fig. 1e and 1g) to follow complete cycles of growth and retraction. During retraction, the GC's height varied between 100 and 200 nm progressively thinning (fig. 3h). After 120 minutes the GC portion indicated by the red arrow (fig. 1f) retracted leaving holes and isolated fragments. The two filaments, possibly cytoskeleton components, initially detected (fig. 1e) completely disappeared (fig. 1h). After 160 minutes the neurite retracted leaving behind it a large fragment isolated from the retracted neurite (indicated by the pink arrow in fig. 1g): no structure with a height larger than 10 nm connecting the fragment to the retracted neurite was detected. We observed the retraction of 7 GCs and in all cases GCs left behind them fragments.
The acquisition of AFM images of living GCs requires several minutes and therefore due to the movement of GCs and to fluidity of the membrane, a high resolution AFM image cannot be achieved. When the same GC is fixed, protein crosslinking increases the stability/rigidity of the membrane and a higher AFM resolution can be obtained.
Effect of Fixation on GC Morphology
With fixed cells, we operated the AFM in contact mode with very low force settings (100 pN-1 nN). In these conditions image resolution is limited only by the radius of curvature of the tip and by the nature of the tip-sample interactions. Therefore by scanning fixed GCs it is possible to resolve details with dimension less than 50 nm. GCs were fixed in situ either with glutaraldehyde and paraformaldehyde after AFM live imaging (fig. 2). In both cases fixation did not modify significantly GCs shape. The height profiles (fig. 2c and 2f) obtained along the blue and the red lines do not superimpose perfectly but GCs volume changed by less than 12%. We compared the structures of 8 GCs before and after fixation and in all GCs the volume was not significantly altered.
Since the high resolution AFM images were obtained only after fixation, we have collected 119 AFM images of fixed GCs from early differentiating ES-derived neurons and we identified 3 GCs morphologies: swollen ad smooth (fig. 3a, c right GC), flat (fig. 3b, c left GC) and fragmented (fig. 3d). The first group represents GCs with several filopodia spreading from the central domain. The height profile always exceeded 200 nm. Flat GCs were highly invaginated with few or any filopodia (fig. 3b and c left GC). Their height profile was consistently below 200 nm. The morphology never reported before is the fragmented one characterized by cellular membrane with several holes (fig. 3d). The height profile of these GCs was also below 200 nm, and their thickness almost vanished in several regions.
Combining DIC, AFM and Confocal Imaging
In order to investigate the dynamics of GC fragmentation we followed with time lapse DIC imaging the motion of live GCs and after fixation we determined their structure with high resolution AFM. The motion of more than 100 GCs was followed by video imaging with time lapse differential interference contrast (DIC) imaging for about 10 minutes. We could categorize 4 different classes of movements: exploring (43/104), growing (16/104), retracting (17/104) and stasis (28/104). 29 of these were also imaged with AFM, followed by staining with appropriate antibodies (fig. 4f,h). Immobile GCs belonging to the last class presented a smooth and flat shape when observed with time lapse DIC images (see fig. 4a, b and e), but their 3D shape was highly fragmented when viewed with AFM (see fig. 4c, g). Immunofluorescence essay (fig. 4f and g) revealed the presence of actin filaments (labeled in blue) on these fragments and which were positive for cellular membrane antibody NCAM (labelled in red).That fragments left behind by retracting GCs are formed by chunks of actin filaments enveloped by the cellular membrane (see registered  confocal image fig. 4h). Moreover AFM profile long the fragments (fig. 4d) shows that their height varies from 50 to 150 nm.
The fragmentation of the cellular membrane was characteristic mostly of isolated GCs. We hypothesized that GCs which had not established physical contact with a suitable target stopped exploring the environment and retracted or degenerated as was previously presented on living and fixed cells (fig. 1 e-g and 4). Indeed time lapse microscopy on living neurons that made contact revealed that their GCs did not move significantly during 10-20 minutes but the surface was smooth and compact on 95% of the cases (data not shown).
Using AFM live imaging we demonstrated that motile GCs had a smooth and intact surface during time (fig. 1a, b and c) while static GCs developed a fragmented shape composed of debris with nanometrical dimensions revealed with AFM as shown in figure 1g and 4c, g. These observations suggest that when in vitro isolated growth cones cannot find the target cell stop their exploration, retract or prune their GC leaving nanoscale debris .
 Alessandrini A. and Facci P.: Sci. Technol. 16 R65 (2005)
 Grzywa E. L. et al.: J. Neurobiology. 66, 1529 (2006)
 Parpura V. et al.: J. Cell. Sci. 104 427 (1993)
 Ricci D. et al.: Methods Mol. Biol. 242, 125 (2004)
 Laishram J. et al.: J. Struct. Biol. 168, 366 (2009)
 Fraichard A. et al.: J Cell Sci. 108, 3181 (1995)
 Keller G.: Genes Dev. 19, 1129 (2005)
 Faulkner R. L. et al.: Dev. Neurosci. 29: 6-13 (2007)
 Franze K., Biophys. J. 97 1883 (2009)
 Ban J. et al.: Stem Cells 25 738 (2007)
Related Articles :
CRN- IOM Laboratorio TASC
Area Scie. Park Basoviz.