Fly strains and crosses were raised on standard cornmeal food at 25°C and on a 12 h light/dark cycle. The following fly stocks were used in this study: neur-GAL4 (Bellaiche et al., 2001), elav-GAL4 (BDSC 458), UAS-H2B-RFP (Mayer et al., 2005), UAS-mCD8::GFP (BDSC 5130), UAS-GMA::GFP (BDSC 31776, referred to as UAS-Actin::GFP), E-Cad::GFP (Huang et al., 2009), Atp-α::GFP (DGGR 110860), Ubi-RFP. nls (BDSC 35496) and Diap1-GFP (Zhang et al., 2008).

Pupae were staged as described in (Bainbridge and Bownes, 1981) and timed employing puparium formation as a reference (hours After Puparium Formation—hAPF). Staged pupae were prepared for live imaging as described in (Mangione and Martin-Blanco, 2020). Briefly, under a dissection microscope, the pupal case was peeled off each pupa using forceps. Naked pupae were transferred to a glass-bottom dish, where each pupa was placed on a small drop of gas-permeable halocarbon oil to reduce refractive index mismatch during imaging with oil immersion objectives. Multiple pupae were mounted for each experiments using this procedure.

In this study, we used inverted Laser Scanning Microscopes (LSM780 and LSM880, Carl Zeiss) equipped with high Numerical Aperture (NA) Objectives (40X/1.3 NA and 63X/1.4 NA oil immersion objective lens). Continuous wave (CW) solid state and Argon lasers were used for the imaging of samples containing GFP and RFP derivatives (laser lines 488 and 561 nm in microscope software, version ZEN 2.3 SP1 FP3 black, Carl Zeiss, Microscopy GmbH). The time interval and number of slices per z-stacks was adapted for each experiment, depending on the size and the developmental stage of the targeted cell. Laser power was kept to a minimum (i.e., typically below 20% for both the acquisition channels) to prevent photobleaching. Pupae were cultured to adulthood after imaging, showing unperturbed development.

We used inverted Laser Scanning Microscopes (LSM780 or LSM880, Carl Zeiss Microscopy Ltd) equipped with Near Infrared (NIR) femtosecond (fs) pulsed laser (Titanium:Sapphire (Ti:Sa) laser; Chameleon Vision II, Coherent Inc), tunable from 680 to 1300 nm for multiphoton imaging. To perform photoablation experiments, the NIR-fs pulsed laser power, pixel dwell time, number of iterations, bleaching ROI size and focus z-position must be precisely controlled. All these parameters were tuned under the ZEN software (ZEN 2.3 SP1 FP3 black, Carl Zeiss Microscopy GmbH, Zeiss, with Bleaching tool tab activated) for each experiment, to ensure successful elimination of the targeted cells. For most of our experiments the MP laser was tuned at 780 nm and set at 70% power, with a pixel dwell time between 1 and 2 μs. Between 1 and 3 iterations were used to ablate circular ROIs ranging between 25 and 35 pixels in diameter (about 1.4 and 2.4 μm). The range of the z-stack was defined in relation to the centre of the stack, to which the NIR-fs pulses were directed (using the z-stack tool tab in centre mode). Following each laser ablation, each z-stack was then acquired in confocal mode to visualize cell behaviours shortly after photoablation.

All images and time-lapses were visualized and processed using Fiji (Schindelin et al., 2012). The Maximum Intensity Projection (Image > Stacks > Z-project) was used to display each z-stack in 2D. The Minimum Intensity Projection (Image > Stacks > Z-project) was used to display bright-field images of the epidermal cuticle. Some z-stacks were displayed in 3D using the 3D viewer option (Image > Stacks > 3D-project) or the Re-slice tool (Image > Stack > Re-slice) of Fiji. Mean fluorescence intensity before and after photobleaching experiments was measured with Fiji (Analyze > Measure > Mean Gray Value). Kymographs from time-lapses were made in Fiji with the Re-slice tool. The NumPy (Harris et al., 2020) package for Python was used for simulating the NIR-fs laser excitation profile, from a Gaussian distribution in the x axis and a Lorentzian distribution for the z axis.

An essential prerequisite to successfully perform photoablation is to label target tissues or structures by the expression of fluorescent markers. Identifying the cell of interest and make its surrounding visible is indeed crucial for precise cellular targeting for photoablation. In this protocol, we use live imaging, so all image acquisition and analysis will depend on genetically encoded fluorescently labelled proteins. When possible, we recommend combining GFP and RFP reporter lines in the same experiments. For example, expression of membrane-localized GFP reporters together with the expression of an RFP-tagged nuclear marker allows the simultaneous visualization of both the geometry and nuclear position of each cell. This combination of cellular markers helps distinguish each cell from its neighbours within the epidermis and associated sensory organs. We also recommend the use of the GAL4/UAS (Brand and Perrimon, 1993) or similar binary systems (Lai and Lee, 2006; Potter et al., 2010; del Valle Rodriguez et al., 2011), to drive the expression of GFP/RFP reporter lines in discrete cellular subsets. For example, the use of the bristle-specific GAL4 driver neuralized (neur-GAL4; see Materials and Equipment) with membrane-localized and/or nuclear fluorescent reporters allows the identification of all bristle cells.

The desired pupal stage must be also selected before photoablation. One of the advantages of the pupa is its immobility, which makes epidermal development and differentiation accessible to live-imaging for over 3 days (Mangione and Martin-Blanco, 2018). Pre-pupal to pupal transitions (i.e., between about 8-to-12 hAPF) and pupal to adult eclosion (i.e., by about 90 hAPF) must be avoided as major morphogenetic movements would hinder imaging. We recommend preparing multiple pupae for each experiment (Mangione and Martin-Blanco, 2020 and Materials and Equipment) so that photoablation experiments can be repeated various times and in multiple animals in each imaging session.

To perform photoablation and live-imaging, we use an inverted laser scanning confocal microscope equipped with a tunable NIR-fs pulsed laser source (Figure 2A; see also Materials and Equipment). Activation of each laser line (i.e., VIS laser module and NIR-fs pulsed laser line), attenuation of laser power for each line, and other acquisition parameters are managed within the microscope software (see Materials and Equipment). For laser scanning, we select the frame mode, and used a 1024 × 1024 pixels image frame size. We recommend setting the scan direction as bi-directional to reduce the scan time (i.e., the duration of the acquisition of the entire frame) during live imaging. For photoablation with the NIR-fs laser, both the dimension and position of the region of interest (ROI) to be ablated within the z-stack are defined using the microscope software (see Materials and Equipment). The number of iterations (the number of scans which are performed for bleaching the selected ROI in each experiment) and scan speed need to be optimized before starting the photoablation (see below). Decreasing the scan speed results in a longer pixel dwell (average time during which the laser beam is focused on each pixel), that can increase the efficiency of bleaching.

NIR-fs lasers emit pulsed radiations that allow the absorption of two or, in principle, more photons in a single quantum event (Denk, 1996; Rubart, 2004). The simultaneous absorption of the two photons induces an electronic transition in the fluorophore from the ground to the excited energy state, equivalent to the fluorescent excitation obtained in single photon mode by using laser lines in the visible range (Helmchen and Denk, 2005). The probability of absorption of multiple photons is non-zero only in a small volume centred in the focal plane where fluorescence excitation takes place. Therefore, the spatial profile of the NIR-fs laser beam is key parameter for predicting targeting precision for photoablation. As the optical resolution of a NIR-fs laser substantially improves in combination with high NA objectives (Rubart, 2004; Vogel et al., 2005), we used 1.3 NA or 1.4 NA objectives in all our experiments. As shown in Figure 2B, the spatial profile of the NIR-fs beam 780–800 nm with 1.3 NA objective shows minimal out-of-focus excitation, an essential requisite for performing highly localized photomanipulations.

Our workflow for photoablation at single cell resolution encompasses a few important steps that ensure successful elimination of the cell of interest, which are described below.

To perform single cell photoablation, we focused the NIR-fs pulsed laser at the centre of nucleus of the targeted cell (Figure 3A), designing a circular ROI using the software of the microscope (see Materials and Equipment). We recommend drawing circular ROIs with diameter no more than half of the nuclear diameter. For example, for a cell nucleus of 5 μm diameter, we recommend designing a ROI of no more than 2 μm of diameter. This strategy helps limiting the extend of thermo-mechanical damage, which expands after laser pulse delivery (Vogel et al., 2005; Vogel et al., 2008; Tsai et al., 2009), to the targeted cell.

Nuclear markers are particularly useful in photoablation experiments, as nuclei between adjoining cells are appreciably separated. In the example provided in Figures 3B,C we marked the nuclei of the epidermal cells with the expression of a Death-associated inhibitor of apoptosis 1—GFP reporter (Diap1-GFP). To precisely deliver high NIR photon density at the centre of the targeted cell volume, we used high NA objective lens (1.4 NA) and high laser power. High NA objectives reduce out-of-focus damage neighbouring cells, while high laser power induce thermo-mechanical damage at the target site. This ensures that the region of highest energy density corresponds to the centre of the cell nucleus (Figures 3A,B), where the photodamage induced by the laser will be confined. Thus, the ROI size and focal plane are important parameters that needs to be adjusted according to the nuclear size/shape of the target cell.

We achieved the best results tuning the NIR-fs pulsed laser at 780 nm but, as laser ablation is mediated through thermo-mechanical damage, the use of a specific wavelength is not critical. Instead, other parameters including laser power, pixel dwell and number of iterations, are all crucial for successful laser ablation within the tissue. We set the NIR-fs laser power at 70% and used a pixel dwell between of 1–2 μs. Together, these parameters allow an efficient photoablation in epidermal cells (Figure 3B; Supplementary Video S1) with a single iteration. We recommend performing a time-lapse imaging experiment post-ablation to verify the extrusion of the targeted cell (Figure 3C; Supplementary Video S2). Taken together, these data show that our system is well suited for photoablation at single cell resolution. The photodamage is here confined to the targeted cell, leaving surrounding cells unaltered. In the next section, we show how we applied this methodology for studying cellular behaviours within tactile bristles or surrounding epidermis in the differentiating PNS.

Notes:

1) Proper alignment in the z-plane of the NIR-fs pulsed laser with the detection beam path is crucial for successful photoablation (or imaging in MP mode). Most microscope manufacture have integrated motorized periscopes and tool tabs to perform this operation.

Each tactile bristle of the Drosophila PNS is innervated by a single bipolar sensory neuron (Hartenstein and Posakony, 1989; Keil, 1997). The dendrite is fully encapsulated by the tactile bristle cells, while the sensory neuron cell body and axon reside basally to the epidermal cuticle, suggesting that dendrites and axons could behave differently after photoablation of the cell body. To test this hypothesis, we highlighted neuronal morphology by expressing a membrane-tagged GFP (UAS-CD8::GFP) under the control of the neuronal driver embryonic lethal abnormal vision (elav)-GAL4 (Figure 4). We then performed targeted photoablation of the neuron cell body (Figure 4A; Supplementary Video S3) and subsequent live imaging to monitor cell behaviour over time (Figure 4B; Supplementary Video S4). The induced photodamage was visible by reduced fluorescent expression of GFP at the site of ablation and fragmentation of the axon, cell body and dendrite (Figures 4B,C), indicating cell death in response to photodamage. Additionally, these experiments further revealed that the dendrite tip persists attached to the bristle, as it retracts after the death of the neuron (Figure 4D). This observation suggests bristle cells protect the distal-most part of dendrite. It should be noted that, in this experiment, we performed successful photoablation using only the membrane marker to direct photodamage to the neuron. In conclusion, the photoablation strategy described here efficiently eliminate sensory neurons in the Drosophila PNS.

All bristle cells originate from a unique precursor cell after four rounds of asymmetric divisions that occur during the pupal stages (Hartenstein and Posakony, 1989; Gho et al., 1996). Upon completion of the lineage, bristle cells enter a postmitotic stage in which each cell executes a complex program of differentiation, to build the adult mechanosensory bristle. For example, the Shaft cell produces the hair shaft, which is deflected in response mechanical stimuli, and the Socket cell builds the cup-like cuticular structure that forms the base of the hair shaft, while contributing to the ionic microenvironment of the tactile organ (Barolo et al., 2000; Jarman, 2002). The Socket and Shaft are sister cells in the SOP lineage, making their individual manipulation particularly challenging. A common outcome of many genetic manipulations at the time of Socket/Shaft cell fate decision is cell fate transformation, leading to adult tactile bristles with two socket-like structures or two hair shaft-like structures (Schweisguth and Posakony, 1994; Barolo et al., 2000). Genetic manipulations that lead to a bristle with one Shaft but no Socket structure and vice versa, are rare or appear at low penetrance, limiting experimental reproducibility. Even with these tools, temporal control of cell elimination is very challenging, in contrast to our method. Moreover, combining those mutant backgrounds with fluorescent reporters for live imaging is often challenging. Photoablation would therefore provide a unique opportunity to elucidate whether a morphologically normal hair socket or hair shaft structure can be formed in complete absence of the corresponding sister cell. Furthermore, the close association of the Shaft and Socket cells during development provides an ideal testing ground to ask if photo-ablation of 1 cell leads to damage in its neighbour. In the example provided in Figure 5, we apply our protocol to photo-ablate the Shaft cell only. We used the bristle-specific driver neur-GAL4 to drive expression of a nuclear marker tagged with RFP (UAS-H2B::RFP) and membrane localized GFP (UAS-Actin::GFP). In this way, all bristle nuclei and cellular structures are visible during and after the experiment (Figure 5A; Supplementary Video S5). In vivo imaging of the differentiating tactile bristle showed that the targeted Shaft cell undergoes cell death, leaving all others bristle cells unaffected (Figures 5B,C; Supplementary Video S6). Furthermore, we inspected the impact of our ablations to the adult organ. We found that the Socket cell developed the cuticular structure typical of the hair base in the absence of the Shaft cell (Figure 5D). This experiment revealed that the presence of the Socket cell alone is sufficient to produce a fully differentiated hair socket in the adult fly.

We next used our protocol to photo-ablate the Socket cell. For this experiment, we marked bristle nuclei (neur-GAL4, UAS-H2B::RFP) and the surrounding epidermal cells with the expression of the Diap1-GFP nuclear reporter (Figure 6). After photoablation of the Socket cell (Figure 6A; Supplementary Video S7), we performed in vivo imaging to follow the response of cells surrounding the photodamaged Socket (Figure 6B; Supplementary Video S8). We observed that, through time, epidermal cells rearrange their positions to seal the space previously occupied by the socket cell, which underwent cell death, while remaining bristle cells continue differentiation (Figure 6B). We then evaluated the impact of Socket cell loss on the adult tactile organ by visualizing the adult epidermal cuticle. We found that the Shaft cell correctly secreted the hair shaft structure and the epidermis surrounding the “Socket-less” bristle appears normal (Figure 6C). This experiment revealed that the presence of the Shaft cell alone is sufficient to produce a fully differentiated hair shaft in the adult fly.

Taken together these experiments show that the Socket and the Shaft cells of the bristle, are relatively autonomous with respect to each other in their differentiation, highlighting a morphological modularity in the making of the adult tactile organ. Furthermore, these data provide a stringent test showing that ablation of one sister in a pair of closely associated cells does not impair the terminal differentiation of the other. Thus, our photoablation strategy is an effective tool to selectively ablate cells within the bristle and will help gaining insights into tactile bristle differentiation.

To further test the spatial precision of our photoablation strategy, we selectively targeted single epidermal cells surrounding the tactile bristle. In the experiment presented in Figure 7, we used photoablation to target a single epidermal cell in contact with the Socket cell. To visualize both epidermal and bristle cells, we used E-Cadherin::GFP (E-Cad::GFP), which localizes to the epithelial adherens junctions, together with the ubiquitous expression of nuclear RFP (Ubi-RFP.nls) (Figure 7A; Supplementary Video S9). After photoablation, the target cell showed reduction in surface area and subsequent extrusion, while surrounding cells rearranged their connections to close the gap (Figure 7B; Supplementary Video S10). This experiment confirms that our method allows localized damage to the targeted cell. Additionally, it makes the targeting of individual epidermal cells feasible, which would otherwise be very labour-intensive to achieve genetically.