Detailed imaging of mitochondrial transport and precise manipulation of mitochondrial function with genetically-encoded photosensitizers in adult Drosophila neurons

Precise distribution of mitochondria is essential for maintaining neuronal homeostasis. Although detailed mechanisms governing the transport of mitochondria have emerged, it is still poorly understood how the regulation of transport is coordinated in space and time within the physiological context of an organism. How alteration in mitochondrial functionality may trigger changes in organellar dynamics also remains unclear in this context. Therefore, the use of genetically-encoded tools to perturb mitochondrial functionality in real time would be desirable. Here we describe methods to interfere with mitochondrial function with high spatiotemporal precision with the use of photosensitisers in vivo in the intact wing nerve of adult Drosophila. We also provide details on how to visualise the transport of mitochondria and to improve the quality of the imaging to attain super-resolution in this tissue.


Introduction
Mitochondria are highly dynamic organelles and their correct distribution is crucial to support many cellular functions. The remarkably long processes of neurons mean that these cells are particularly dependent on mechanisms for long-range cytoskeletal transport of mitochondria. Due to their peculiar architecture, and the stereotypical directionality of the transport in the axons, neurons are a particularly good model to understand the logic of intracellular tra cking.
The fundamental importance of axonal transport for achieving maturation and maintaining the functionality of neurons is well documented [1] and many models have been developed to visualise this process. High throughput, ease of gene manipulation and accessibility to pharmacological treatments make cultured neuronal cells, for example primary neurons or neurons derived from neuroblastoma and stem cells, arguably the mainstay of neuronal tra cking studies and have signi cantly contributed towards the mechanistic understanding of the transport process [2][3][4][5][6][7][8][9][10][11][12][13]. Ex vivo models, where a whole tissue is dissected and maintained in culture or imaged directly after dissection, are valid alternatives [14][15][16][17][18][19][20] with the advantage of potentially preserving a near-native environment while affording accessibility for pharmacological and electrophysiological studies.
In vivo animal models to study the transport process in neurons are now also widely used [6, 21-29]. A clear advantage of an in vivo system is the possibility of studying the transport process while retaining the full complexity of an organismal setting, which may be preferable when studying axonal transport during ageing and in animal models of age-dependent neurodegenerative disorders [30]. However, because the neurons of interest are often buried deep into the tissue that is being imaged, this advantage tends to be offset by the need of complex surgery required to visualize the neurons of interests. This is the case, for example, for the implantation of cranial windows for the observation of thalamo-cortical projections in mice [31] or for the exposure of the mouse sciatic nerve for tra cking studies [32]. In addition, the resolution at which dynamic subcellular components are resolved during in vivo imaging is often modest compared to the quality that can be obtained from in vitro cultured cells. Despite the current rapid development of super-resolution techniques, only few in vitro studies have focused on long-range intracellular transport captured by live-cell super-resolution imaging [33,34]. While challenging, achieving super-resolution of tra cked organelles in vivo would be desirable, and a tangible step forward, to better understand the intricacies of intracellular tra cking in this context.
Performing gene manipulations to interfere with the transport or function of speci c cargoes might not be straightforward in vivo, especially in mammalian systems. Knockdown or knockout studies are timeconsuming; acute modulation of transport, for instance by the addition of drugs, also presents challenges in an in vivo setting, limiting the possibility of gaining in-depth mechanistic insight of this process. This is problematic when studying mitochondrial dynamics which often relies on pharmacological manipulation to interfere with different stages of the oxidative phosphorylation. We developed a system for detailed imaging of organelle transport in adult Drosophila in vivo, which exploits the accessibility of the wing neurons for microscopic observation and does not require surgical procedures [35]. In this chapter, we describe methods to image mitochondrial transport in different neuronal population of the wing using anaesthetized adult animals. We have previously showed the suitability of this tissue for structured illumination microscopy (SIM) of neuronal membranes and nuclear markers [35]. With the aim of improving the resolution for in vivo imaging of mitochondrial tra cking, here we describe the application of Super-Resolution Radial Fluctuation (SRRF) microscopy [36, 37] to live imaging in the wing nerve of adult Drosophila. Finally, we report the generation of new transgenic ies encoding the photosensitisers KillerRed and SuperNova targeted to the mitochondria and describe photostimulation protocols to induce mitochondrial damage rapidly, coupled to the imaging of the redox state and motility of the organelles. We believe this will expand the utility of the wing nerve system and provide additional tools for the study of mitochondrial function and transport in vivo. for kymographs generation (see Note 6).

Microsoft Excel and GraphPad
Prism to analyse and plot the data.

Methods
We use the GAL4-UAS binary system [39] to drive the expression of GFP in the mitochondria of the adult wing neurons. By using either a pan-neuronal driver, for example nSyb-Gal4, or a more restricted driver such as the cholinergic neuronal driver ChAT-Gal4, mitochondria can be visualised clearly in the neurons of the adult Drosophila wing ( Figure 1B-D) (see Note 7). However, application of the SRRF algorithm to time series acquired by spinning disk microscopy allows to achieve a considerably improved resolution of both stationary and moving mitochondria ( Figure 1D and Movie 1), which is useful to appreciate ner details of mitochondrial dynamics.
In order to damage mitochondria with spatiotemporal precision, we produced novel transgenic Drosophila lines expressing the photosensitisers KillerRed (KR) [40] and SuperNova (SN) [41] targeted to the mitochondrial matrix (mito::KR and mito::SN, respectively) (Table 1, Figure 2 and Figure 5A). Upon irradiation with green light, KR and SN produce a large amount of reactive oxygen species (ROS). As direct readout of the e cacy of our stimulation protocol, we measured the mitochondrial redox state with the roGFP2-Grx1 reporter (mito::roGFP2-Grx1), which exhibits uorescence emission shift from 488 nm to 405 nm in more oxidised mitochondrial environments [42,43]. Brief stimulations of mito::KR with a 561nm laser light are su cient to trigger a rapid and transient increase in mitochondrial oxidation ( Figure 3).
Our protocol allows to stimulate areas covering a large portion of the nerve ( Figure 3) as well as achieving precise photostimulation of a single mitochondrion ( Figure 4). Mito::KRneurons that do not express mito::KR do not show a detectable mitochondrial oxidative response when subjected to the same photostimulation regime, ruling out the possibility that the exposure to the 561-nm laser light alone is su cient to strongly increase mitochondrial oxidation, at least for the brief irradiation times used here.
No signi cant change in the mitochondrial redox state was observed after brief stimulation of mito::SN with a 561-nm laser light ( Figure 5B-D). This suggests heightened sensitivity to mito::KR in our system which we nd advantageous to be able to probe the early consequences of mitochondrial oxidation while minimising light-induced phototoxicity. SuperNova is a monomeric variant of KillerRed and was shown to generate a different proportion of singlet oxygen and superoxide species compared to KillerRed [41,44].
This was not reported to diminish its phototoxic effect [41], although a robust effect of SuperNova stimulation might become clearer on longer timescales in vivo or after longer irradiation times [45]. SuperNova2, a variant of SuperNova with enhanced phototoxicity has been recently engineered [46] and could be a valid alternative to enhance the phototoxicity of the protein in an in vivo setting.
To investigate the consequences of acute mitochondrial damage on the motility of the organelle, we present two specular approaches. In the rst approach, KillerRed is bleached in mito::KR mito::GFP neurons and the damaged mitochondria undergo a yellow-to-green colour switch ( Figure 6A and Movie 2). An alternative approach is to use mito::KR in combination with a mitochondrially-targeted photoactivatable GFP (mito::PA-GFP). In this case, simultaneous stimulation with 405-nm and 561-nm lasers leads to quenching of the red signal and concomitant activation of the green uorophore, with mitochondria undergoing a red to green transition ( Figure 7A and Movie 3). Co-expression of the two mitochondrial markers does not affect the motility and morphology of the organelles and, with both approaches, irradiated (i.e., dysfunctional) mitochondria can be tracked through colour switch. However, the use of PA-GFP is preferable when the aim is to study the dynamics of the irradiated mitochondria, as only one laser line (488 nm) is required to directly follow the mitochondria after photostimulation. To study the dynamics of both irradiated and non-irradiated mitochondria, the simultaneous use of both 488-nm and 561-nm laser lines for the whole duration of image acquisition is recommended.
3.1 Generation of plasmids for the production of transgenic lines 1. The Cox VIII MTS, KillerRed and SuperNova are PCR ampli ed using the Q5® Hot-Start Fidelity 2X Master Mix (NEB) and FWD and RVS primers reported in section 2.1, which include overlapping sequences complementary to adjacent fragments and necessary for the assembly as described in step 3.
2. The PCR products are puri ed through solid-phase puri cation using spin columns (we routinely use Monarch® PCR&DNA Cleanup kit, NEB).

The NEBuilder®
HiFi assembly method is used to introduce the KillerRed and SuperNova in the pUASP vector, downstream of two MTS from the human cytochrome c oxidase subunit VIII (see Note 8).

Fly maintenance
1. Cross males and virgin females to obtain ies of the desired genotype. In our protocol, the Gal4 drivers and UAS responders are heterozygous in all genotypes.
3. Remove the ies after 4-5 days and keep the vial with wandering larvae.
4. Collect newly emerged ies and select the desired genotype. Put the selected ies in new vials, maintain them at 25°C with a 12h-light/12h-dark cycle and image them after 24-48h, when approximately 30% of mitochondria are motile [47].

Sample preparation
1. Anesthetise the ies on a CO 2 pad (see Note 9), while preparing the imaging chamber ( Figure 1A).
2. On each of the short sides of the coverslip, make a stack of three layers of masking tape.
3. Add a piece of double-sided tape in the middle of the coverslip and on each stack of masking tape. 4. Apply a thin layer of 10S halocarbon oil on one side of the double-sided tape in the middle of the coverslip.
5. Mount one y on the coverslip. Pick the y up by the legs and place it on the oil, with the ventral side up and with the head on the double-sided tape in the centre of the coverslip.
. Dip a ne paintbrush in oil and use it to spread out the wings. Make sure that the wings are at on the coverslip and most air bubbles are removed (see Note 10).
7. Place a second coverslip on top to seal the chamber and image the y immediately. To minimise delay, we recommend mounting the y in proximity of the microscope used for live imaging.
3.4 Imaging axonal transport in the Drosophila wing 1. Set the microscope temperature controller at 25°2 . Mount the ies expressing a uorescence marker (for instance, GFP) targeted to neuronal mitochondria.
3. Find the region of interest in the Drosophila wing by using the transmitted light (through eyepieces).
4. Inspect the whole wing for damage (for instance, small cuts). If the wing is damaged, discard the chamber and mount another y.
5. Switch to uorescence mode to check the uorescent signal (through eyepieces) and then to confocal mode (through microscope software) to acquire the images.
. Adjust exposure time and laser power to obtain a good signal, clearly visible above background.
2. Adjust exposure time and laser power to obtain a good signal, clearly visible above background.
Here, we used 30 ms exposure time and 50-60% laser power (see Note 12).
3. Change the size of the eld of view so that the time to acquire a single frame is shorter than the chosen exposure time (see Note 13).
4. Acquire a series of 33 frames by continuous imaging, i.e., where the interval between each frame is limited by the exposure time, in this case 30 ms. Each series of 33 frames will be used to create a single SRRF timepoint (see section 3.6).  3.8 Photostimulation in the Drosophila wing: assessing mitochondria motility upon acute mitochondrial damage . Run "Ratio pro ler" (Cookbook > Image Intensity Processing > Ratio Pro ler). A "Results" tab will open and report the ratio between Channel 1 and Channel 2 for each frame. This value is the 405 nm / 488 nm ratio for the mito::roGFP2-Grx1 probe, used to quantify the mitochondrial redox state.
. To control for the speci city of the irradiation to the region de ned by the ROI, run "Ratio pro ler" again on an area that has not been stimulated ( Figure 3C . Press the "k" button to generate the kymograph for each channel. 7. Merge the two kymographs ( Figure 6C and 7C).

Notes
1. Flies were grown on 'Iberian' food, although other recipes are also suitable for visualisation of cargo transport in wing neurons. It is crucial, however, to be consistent with the recipe used as changes in the diet can in uence the outcome of the experiments. We found that inconsistent use of the yeast and our can be the source of experimental variability.
2. To minimise variability in the experimental outcome, it is important that temperature, humidity and light-dark cycle are maintained constant.
3. The shorter working distance afforded by the coverslips n. 0 means that acquiring clear images is straightforward. However, these coverslips are thin and very exible, which makes stabilising the sample in the imaging chamber more di cult and might lead to drifting while imaging. We nd coverslips n. 1 to be a good compromise between quality of images obtained, sample stability and maintenance of an appropriate optical con guration. However, coverslips n.1.5 are recommended for live super-resolution imaging (e.g., instant SIM). 4. We compared the 10S halocarbon oil from both VWR (cat. no. 24627.188) and the halocarbon oil 700 from Sigma Aldrich (cat. no H8898). We found the former to be more viscous and to better aid sample stabilisation. However, using the Sigma oil does not prevent obtaining high-quality images.
5. Different microscopes can be used to this purpose. We nd spinning disk confocal to be the best option for live imaging, as it is gentler on the sample (i.e., less phototoxicity) and affords both highresolution and high-speed. We routinely use two types of confocal settings: a Nikon Ti-E inverted system equipped with a Yokogawa-CSU-X1 Spinning Disk head, Andor iXon Ultra DU-897 EMCCD Camera (for axonal transport studies and SRRF) and an inverted Nikon A1RHD confocal microscope equipped with both GaAsP, multi-alkali and spectral detectors equipped with a photostimulation unit (for light-induced activation of the photosensitisers). With both systems, imaging was performed with a Nikon 60X/1.4NA Plan Apochromatic oil immersion objective.
. Other kymograph analysis bundles can also be used effectively in this system, for example, KymoClear and KymoDirect [48,49], KymoAnalyzer [50] or KymoButler [51]. Alternatively, a kymograph of only motile mitochondria could be generated from manually annotated tracks. This would eliminate the background associated with the numerous stationary mitochondria in the wing neurons.
7. The ChAT-Gal4 driver can be used effectively to visualise mitochondria in the wing neurons using different uorophores. The mitochondria in ChAT-Gal4 + neurons display robust motility and invariably good uorescent signal in both the L1 and L3 wing veins. ChAT-Gal4 is therefore an excellent alternative to pan-neuronal and more restricted drivers currently used in this system for live imaging of axonal transport [35,[52][53][54][55]. We do not recommend the use of ubiquitous drivers as the uorescent signal from non-neuronal cells in the wing would mask the signal from the axons of the marginal nerve.
. We found that fusing KillerRed and SuperNova to one MTS results in diffuse cytoplasmic distribution of the proteins in Drosophila S2R+ cells, while using two MTS is su cient for e cient mitochondrial targeting. 9. Make sure that the selected y does not carry markers that affect wings morphology (e.g., CyO).
Carefully check the wings at the stereo microscope. If damaged, discard the y.

10.
Wings not being at on the coverslip will decrease the precision of the photostimulation and the clarity of the signal. The wing arch region (which is closer to the body) might not be at on the bottom coverslip as a result of head and the thorax of the y becoming unstuck from the coverslip, for instance if the head of the animal is not well attached on the double-sided tape. In this case, the application of little oil could be used to further restrain the animal. We found that ner round tip paintbrushes (for example, n.1) are particularly suitable to carefully position the y.
11. It is possible to image mitochondrial transport in the wing for up to 30 minutes [35]. However, when imaging for longer periods of time, we recommend using a slower frame rate (e.g. 0.5 or 0.33 fps) to reduce phototoxicity. On the other hand, if the study focuses on the analysis of instantaneous velocities, recording at higher frame rates for a shorter period is more appropriate.
12. For good quality SRRF movies, a strong uorescent signal and acquisition at high fps are essential. Therefore, we recommend against using dim uorophores. 13. To obtain a single SRRF image, a series of frames needs to be acquired in a short time interval (e.g. 1 s) and exposure times between 10 ms and 100 ms are recommended [37].
14. The effective elapsed time between time series is de ned by the combination of both the time needed to acquire 1 loop of 33 consecutive frames (1 second, with our settings) and the interval between each loop (2 seconds in our case). Our effective acquisition time is therefore of 0.33 fps. In an imaging period of 90 s, this equals to 30 acquisition loops for a total of 990 frames. These settings allow to follow mitochondria reliably without signi cant photobleaching.
15. The settings were chosen by adjustment and comparison of several images. Ring Radius, used to calculate the intensity gradients of nearby subpixels, was chosen by iterative trials of different radii, until an optimal response of the analysis was observed, typically judged by the best resolution improvement and minimal patterning. Radiality Magni cation, which determines the number of subpixels each pixel is split into, was not increased to more than 2 as improvements were not evident at greater values. 17. At constant laser power, high scanning speed and high number of loops result in more precise bleaching than a low scanning speed and low number of loops. For experimental design, it is useful to consider the time it takes to switch between optical con gurations or stimulation loops, which may vary considerably depending on the microscope setup.
1 . Our stimulation protocol is su cient to trigger a rapid change in the redox state of the mitochondria. However, this recovers almost to pre-stimulation levels within 5 min ( Figure 3C-D). If desired, longer and more frequent stimulation loops can be used which may result in sustained oxidation of the irradiated area. In this case, however, it is important to carefully control for potential light-induced mitochondrial oxidation.
19. The size of the ROI can be chosen depending on the speci c experimental aim. To bleach moving mitochondria, the bigger the stimulation ROI the higher the probability that motile mitochondria will cross it at any given time. In the wing neurons, average mitochondria speed is approximately 0.2 µm/s and 0.4 µm/s for anterograde and retrograde transport, respectively [47]. With a stimulation step of 10-15 s, the displacement of irradiated motile mitochondria could be of up to approximately 2-6 µ We found that using a 15-25-µm-long stimulation ROI is su cient for reliable bleaching of mito::KR in motile mitochondria.
20. Ratiometric images were generated with the "RatioView" modality of NIS-element software. Alternatively, the Image Calculator command in Fiji can also be used.
21. Some areas immediately outside the bleached ROI are likely to be reached by the scattering of the excitation light and may respond to the stimulation protocol. This can be ascertained empirically and the extent of the irradiated sample outside the boundaries of the ROI minimised. When quantifying control regions (i.e., areas that have not been stimulated), it is therefore preferable to choose an area not adjacent to the irradiated ROI. Alternatively, using two-photon microscopy to perform the photostimulation experiments is likely to reduce the scattering of light in the tissue.        Assessing mitochondrial motility after acute mitochondrial damage using mito::KR and mito::GFP. (A) The cartoon illustrates the experimental design. After bleaching KillerRed with the 561-nm laser, mitochondria undergo a yellow-to-green transition, indicated by the change in the colour of the mitochondrial matrix in the cartoon: the red KillerRed signal is quenched and only the green GFP uorescence is retained in the mitochondria. Bleached (i.e., dysfunctional) mitochondria can therefore be followed by illumination with both 488-nm and 561-nm lasers. (B) Neurons in the margin of the wing nerve before and after irradiation (dashed black line). The white rectangles indicate the photostimulated ROI. The area shaded in the bottom panels was used to generate the kymograph in (C). Scale bar: 10 µm.