Tracking neuronal motility in live murine retinal explants

Summary The developing retina undergoes dynamic organizational changes involving significant intra-retinal motility of the encompassing cells. Here, we present a protocol for tracking retinal cell motility in live explanted mouse retinae. Although originally applied to rod and cone photoreceptors, this strategy is applicable to any fluorescently labeled cell in mouse retinae and other similar experimental retinal models. Careful tissue handling is critical for the successful acquisition of high-quality live imaging data. Further instructions for semi-automated in silico data handling are provided. For complete details on the use and execution of this protocol, please refer to Aghaizu et al. (2021).


SUMMARY
The developing retina undergoes dynamic organizational changes involving significant intra-retinal motility of the encompassing cells. Here, we present a protocol for tracking retinal cell motility in live explanted mouse retinae. Although originally applied to rod and cone photoreceptors, this strategy is applicable to any fluorescently labeled cell in mouse retinae and other similar experimental retinal models. Careful tissue handling is critical for the successful acquisition of high-quality live imaging data. Further instructions for semi-automated in silico data handling are provided. For complete details on the use and execution of this protocol, please refer to Aghaizu et al. (2021).

BEFORE YOU BEGIN
This protocol was designed to track motions of fluorescently labeled retinal cells. In Aghaizu et al. (2021), cellular labeling was achieved either by using transgenic mouse lines expressing cell-type specific fluorescent reporters or by electroporation of acutely explanted retinae with an expression plasmid driving the expression of a fluorescent protein.
All mice in preparation of this manuscript were used according to the NC3R ARRIVE guidelines. Both male and female mice were used in this study without discrimination. All animal studies were carried out under the Animals (Scientific Procedures) Act 1986 under a project license PPL 70/8120 issued by the UK Government Home Office and conducted in accordance with protocols approved by the Animal Welfare and Ethics Committee of the UCL Institute of Ophthalmology. All animals were killed by cervical dislocation performed by trained personnel (approved under Schedule 1 as a method of humane killing). All efforts were made to minimize the number and suffering of animals used in these experiments.
Male and female mice were group housed in the animal facility at University College London on a standard 12-h light/dark cycle at the same light levels throughout the experimental period. Animals were kept in individually ventilated cages on animal grade wood chips and given access to nesting material and food and water ad libitum.
1. Cell-type specific reporter mice are used for the purpose of having fluorescently labeled rod or cone photoreceptors in vivo. Timed-matings need to be set up 2.5-4.5 weeks prior to retinal live imaging depending on the developmental stage of interest (typically between embryonic day (E) 16.5 and postnatal day (P) 10.5). a. For fluorescently labeled rod photoreceptors, time-mate Nrl.GFP +/+ mice (Akimoto et al., 2006). These mice are suitable for homozygotic breeding. b. For fluorescently labeled cone photoreceptors, time-mate Chrnb4.EGFP mice (Gong et al., 2003). These mice should be bred as hemizygotes. c. For Cre-expressing rod photoreceptors, time-mate Nrl.Cre +/À mice (Brightman et al., 2016).
These mice should be bred as hemizygotes. 2. Successful mating is typically accompanied by the presence of a vaginal plug, which marks embryonic day (E) 0.5 following overnight mouse pairing for 14-16 h. The gestation period in mice is usually 19-21 days (strain dependent). Subsequently isolate and live image retinae at desired time point (see 'step-by-step method details').

Preparation of transfection-based cellular labelling
Timing: 2.5-4.5 weeks including husbandry; 2.5 h for dissection, electroporation and culture setup 3. Photoreceptor fluorescent labeling can also be achieved by introducing an expression plasmid of choice into the desired target retinal cells. We used a conditional strategy to limit fluorescent protein expression to rod photoreceptors only. This was achieved by electroporating a Cre-sensitive expression plasmid into acutely explanted retinae from Nrl.Cre +/À mice as described below, although this can also be achieved by in vivo electroporation (Matsuda and Cepko, 2004) or AAV viral transduction (Aghaizu et al., 2021). a. Dissect out retinae from P0.5-1.5 mice obtained from timed mated pregnancies (see 'preparation of transgenic mice'; (1.c)) in ice cold dissection medium ( Figure 1A). Remove all non-neural retinal tissues (cornea, sclera, iris, choroid, optic nerve, retinal pigment epithelium, and ciliary body) with exception of the lens, which is left in place for increased tissue stability ( Figure 1B).
CRITICAL: To avoid contamination, dissection tools (#5 Dumont forceps, Vannas scissors) are to be sterilized with 70 % ethanol before dissection. Furthermore, aim to perform the majority of 'Preparation steps no. 3' in a sterile tissue culture hood.
CRITICAL: Care should be taken to avoid damaging retinae during dissection as tissue disruptions introduced during this step tend to become amplified during in vitro culture. We found that dissected retinae are most gently transferred by using sterile plastic Pasteur pipettes (supplied with electroporation cuvettes) with a manually widened opening ( Figure 1I).
b. Transfer retinae to electroporation cuvette containing a 1 mg/mL plasmid DNA solution in PBS, prepared as described in 'Materials and equipment' (e.g., shDync1h1/EGFP; Aghaizu et al., 2021) and incubate at room temperature (18 C-21 C) for 5 min. c. Orientate retinae within the cuvette such that the apical and lens sides of the retina are aligned with the cuvette electrodes; for optimal electroporation efficiency the apical surface is to be juxtaposed to the anode ( Figure 1J). d. Use a pulse generator to electroporate retinae with 5 3 30 V square pulses of 50 ms duration and with 950 ms intervals.
CRITICAL: Avoid direct contact between electrodes and retinae during electroporation as this may result in substantial tissue damage.
ll OPEN ACCESS f. Transfer retinae to pre-equilibrated (37 C/5% CO 2 ) culture medium for 5 min. g. For in vitro retinal explant cultures, transfer up to 3 retinae lens side down onto a Whatmanâ Nucleporeä Track-Etched Membrane floating on 2 mL of culture medium in a 6-well tissue culture plate and culture at 37 C/5% CO 2 in a sitting drop of culture medium; 1 retina per drop ( Figure 1J). h. Re-feed each retina by adding 25 mL culture medium to sitting drop every other day until ready for live imaging.
Note: Transfection efficiency can be assessed on standard epifluorescence microscopes, fluorescent protein permitting. In general, latency times between electroporation and visual fluorescent protein detection depend to a large extent, albeit not exclusively, on promoter strength and fluorescent protein brightness. In our experience, when expression of EGFP and DsRed was driven by a strong promoter (e.g. CMV, CAG or rod cell-specific Nrl promoter (Akimoto et al., 2006)), expression was detectable after 1-2 days and firmly established at 3 days post electroporation.

MATERIALS AND EQUIPMENT
Note that highly concentrated plasmid DNA (ideally > 1 mg/mL) is required and should be produced using Qiagen plasmid kits or similar commercially available kits. Dissection medium was freshly prepared prior to retinal harvest but should be stable at 4 C for 1 week.

Total N/A 10 mL
Culture medium can be stored at 4 C for 1 week.

OPEN ACCESS
Note that rutin is supplied as part of DMEM gfp -2 kit (Cat#: MCK02). ii. Enucleate eyes and transfer to ice-cold RPMI 1640 medium using forceps ( Figure 1A). iii. Isolate neural retina (tissue composed of cells of the outer nuclear layer, inner nuclear layer, and ganglion cell layer) by dissection under a dissection microscope in ice-cold RPMI 1640 medium by removing all non-neural tissues (cornea, sclera, iris, choroid, lens, optic nerve, retinal pigment epithelium, and ciliary body) using forceps (Figures 1B and 1C).

STEP-BY-STEP METHOD DETAILS
CRITICAL: Upon removal of the lens during retinal dissection, ensure that vitreous is removed as well. Failure to remove vitreous can impede the flattening process (also applies to 1.b.ii).
CRITICAL: Care must be taken to remove all pigmented non-neural retinal tissues (retinal pigment epithelium and ciliary body) as the melanin pigment contained in these tissues absorbs the 2-photon infrared light leading to rapid heat-induced tissue damage when the microscope is operated in 2-photon mode (also applied to 1.b.ii). 2. Flattening and mounting retina. a. Apply four equidistant radial cuts using fine scissors going halfway towards the centre of the retina. This results in 'petal' like retinal segments ( Figure 1D). b. Transfer retina in a small drop of RPMI 1640 medium onto the edge portion of a microscopy glass slide using a sterile plastic Pasteur pipette with manually widened opening ( Figure 1I) such that photoreceptors face the glass slide and retinal ganglion cells face the user ( Figure 1D).

Reagent Final concentration Amount
Rutin (1003)  c. Carefully flatten retina onto glass slide by unfolding petals using fine forceps. Petals will resist flattening due to natural curvature of the retina. Flattening is aided by removing some of the RPMI 1640 ( Figure 1E).
CRITICAL: Care must be taken to not introduce tissue damage during flattening and unfolding petals. This may be aided by using a clean, fine brush. Even small tissue disruption may result in visible loss of retinal integrity during live imaging.

OPEN ACCESS
CRITICAL: (2.c.-3.b.) During these steps, the retina will be exposed to minimal amounts of RPMI 1640 medium at times. Care must be taken to not let the retina dry out.
d. After successful flattening, remove most of RPMI 1640 medium using lint-free tissue paper. e. Immediately place a 0.45 mm MF-Milliporeä nitrocellulose membrane on the retina (with ganglion cell layer touching the membrane) making sure tissue is securely mounted on the membrane ( Figure 1F). f. Gently transfer membrane-mounted retina back into RPMI 1640 medium by dipping the edge of the glass slide into medium ( Figure 1G). The membrane will gradually become visibly hydrated when contacted by the solution, upon completion of which the membrane can be fully transferred using the dissection tools ( Figure 1H).
3. Setting up imaging dish for live imaging (Figure 2) a. Fix membrane with retina facing up to the bottom of a Ø5.5 cm tissue culture petri dish on a thin layer of vacuum grease. b. To ensure secure mounting and to prevent tissue drifting during live imaging, carefully press down membrane onto vacuum grease using blunt tools. Place a platinum ring over the membrane, the added weight of which should further prevent tissue drifting. c. Fill the dish with 10 mL DMEM gfp -2 live imaging medium.

Live imaging setup and recording
Timing: 45 min set up; up to $ 12 h time-lapse recording The following steps describe setting up the retinal live imaging session (see also Figure 2), at the end of which the user will have generated a time-lapse recording of the fluorescently labelled retinal cells. We routinely used a 2-photon microscope set-up to reduce phototoxicity and to increase photon tissue penetration and z resolution.
CRITICAL: Allow tissue to acclimatize to 37 C/5% CO 2 for at least 30 min prior to initiating time-lapse recording to prevent temperature gradient induced tissue drifting. 5. Carefully introduce a water-immersion objective pre-sterilized with 70 % ethanol into the live imaging medium. We routinely used 253 and 403 water-immersion objectives (NA = 0.95 and 0.8 respectively). 6. Affix CO 2 supplying apparatus to the live imaging dish. 7. Identify suitable field of view to be subjected to time-lapse recording by scanning through microscope eyepiece. 8. Tune the multiphoton laser source to a wavelength suitable for 2 photon-excitation of the fluorescent protein(s) of interest (e.g., 920 nm for EGFP). 9. Further, ensure that laser and detector(s) have been properly set up for signal detection.
Note: In 2-photon mode, the pinhole of confocal microscopes should be fully dilated.
10. Apply a layer of paraffin oil, pre-equilibrated to 37 C, on top of the live imaging medium to prevent medium evaporation during time-lapse recording. 11. Initiate time-lapse recording. In Aghaizu et al. (2021), we acquired xyzt image series at a resolution of 512 3 512 pixels, at a step size of 1 mm and at 10 min intervals. Imaging duration frequently reached up to 12 h, although this is critically dependent on starting tissue health/ integrity.
CRITICAL: Any pigmented cells present within the field of view will absorb infrared light (i.e. 2-photon laser light) and rapidly heat up the tissue, rapidly bringing about tissue destruction.
Note: It is recommended to ensure on occasion that the intended field of view remains in focus; in the event of tissue drift appropriate adjustments are to be made by user.
Note: When imaging photoreceptors, early stages of deteriorating tissue health most frequently presented themselves as individual photoreceptors apically extending out of the outer nuclear layer (see Figure 4A, bottom right panel for an example of apically extruding photoreceptors). Further deterioration usually entails full apical 'ejection' of individual and subsequently bigger groups of photoreceptors until tissue integrity is fully compromised.
In silico processing of time-lapse recording and cell tracking Timing: minimum 30 mins but can be significantly longer depending on image quality, number of cells to be tracked, and implementation of manual track validation.
Upon completion of time-lapse recording, generated xyzt image series are to be processed and subjected to cell tracking.
12. Process xyzt image series in Fiji/ImageJ (Schindelin et al., 2012) with standard imaging processing tools available, such as noise reduction, to achieve maximum possible image quality and signal to noise ratio. Ensure that processing steps are kept consistent. 13. Register xyzt image series (in silico correction of tissue drift), as some amount of tissue drifting during time-lapse recording is common. This can be accomplished in Fiji/ImageJ either with the 'Correct 3D drift' plug-in (Parslow et al., 2014) for automated drift correction based on automated landmark recognition or the 'Descriptor-based series registration (2D/3D + t)' plug-in (Preibisch et al., 2010) in cases where automated registration failed and registration landmarks had to be defined manually. 14. Export registered xyzt image series as *.tif file. 15. Perform cell tracking using software of choice. In Aghaizu et al. (2021) we used the spot tracking tool within IMARIS software. IMARIS software is able to identify rod cell bodies based on the ll OPEN ACCESS preceding input of size and location parameters (Methods video S1; video still shown in Figure 3). Rod somata were typically ellipsoid in shape, measuring $ 5 mm along the short (lateral) and $ 7.5 mm along the long (radial) axis and are typically located deeper within the tissue compared to the developing bulbous segment structures.
Note: Once set up, cell tracking in the IMARIS spot tracking tool is fully automated. Nonetheless, we recommend manually validating the generated trajectories to avoid including data artifacts into subsequent analysis steps.
16. Export completed cell tracking data as *.xls file.

Data processing and analysis
The primary data outputs from the processing and analysis steps will be average mean squared displacement (AvMSD) as well as instantaneous velocity and instantaneous acceleration values for the population of cells analysed. In Aghaizu et al. (2021), these were generated by importing IMARIS-generated tracking information into Matlab followed by execution of custom-written Matlab scripts (available on the project Github repository on https://github.com/RPearsonLab/ Photoreceptor_tracking).
17. Arrange IMARIS-generated tracking data in Microsoft Excel according to Table 1. '#' denotes numbered data point ID in Excel; 'Trajectory ID' denotes unique numbered identifier of individual trajectories; 'Time index' denotes time point within time-lapse recording, at which data point was generated (note that this is in integer rather than time units); 'Position x/y/z' denotes 3D positional information.
Note: Matlab scripts were designed to read-in tracking data in the format specified in Table 1. The user is advised to either manipulate data to match Table 1 or to modify Matlab script accordingly.
18. Open and run the Matlab script 'MSDforztrackingdata.m' (see Github project repository) to read-in trajectory data and to calculate AvMSD & standard error values. Users will be prompted to navigate to the source trajectory file upon initiation of the script. AvMSD will be calculated according to the equation ½zðði + nÞdtÞ À zðidtÞ 2 where zðidtÞ and zðði + nÞdtÞ are the z positions of a given nucleus at time frames idt and ði + nÞdt respectively, n is an integer representing the time interval between those positions and N is the total number of time points within the time-lapse recording. The only script manipulation required to run 'MSDforztrackingdata.m' is to specify the range of trajectory IDs to be analysed from source file (e.g.,: 1, 2, 3,., n); line 134. 19. Upon completion of script 'MSDforztrackingdata.m', the Matlab will have generated result outputs in Vector/matrix format denoted 'AvMSD' and 'standarderrors' containing AvMSD and standard error values with increasing time lag.
Note: For our cell type of interest (photoreceptors), motility occurred predominantly along the z axis (radial) rather than in xy (lateral) directions. AvMSD and instantaneous velocity calculations in the Matlab script are thus based exclusively on z motility and the user is advised to adjust the Matlab script accordingly for higher dimensional analyses.
20. Open and run 'ztrajectory_gen.m' to convert read-in trajectory data stored in array format into matrix 'ztrajectories'. Individual trajectories are listed by row.  Note: Matlab script was custom written for time-lapse recordings acquired at 10 min intervals. This is reflected in instantaneous velocity calculations, where velocity between consecutive frames (mm/10 min) needed to be divided by 10 to obtain instantaneous velocity (mm/min) (lines 59 and 64 in 'Trajectory_analysis.m').

EXPECTED OUTCOMES
Following successful live imaging the user will have generated a xyzt image stack. The fluorescence signal distribution within the tissue will depend on the exact experimental design. For reference, we provide two examples where rod photoreceptors were labeled using different approaches (Figure 4A): Nrl.Cre +/À retina, electroporated at P0.5 with shDync1h1/EGFP and cultured for 6 days in vitro (DIV) before live imaging (left panels); Nrl.GFP, postnatal day (P) 7.5 (right panels). Rod somal position can then be tracked as described in step 15. Top-down (xy) and orthogonal views (xz; maximum z projection of 5 adjacent slices at 1 mm interval) of the image stacks are shown. For the purpose of this protocol, the presented quantifications are based on the example trajectory data provided in Table 1.
Changes in z position over time were plotted for the example trajectory data ( Figure 4B). Average mean squared displacement and standard error values were obtained using the provided Matlab script (step 18.) and plotted versus time lag ( Figure 4C). Note the positive curvature in the data points, due to which quadratic function curve fitting (y = ax 2 ) should be applied to obtain the quadratic coefficient. Where average MSD data points follow a straight line, linear curve fitting should be applied, in order to obtain the coefficient of movement according to step 22.a.i. Instantaneous velocity values were derived using the provided Matlab script (steps 21, 22.) and plotted as a histogram ( Figure 4D).

LIMITATIONS
The protocol described here is inherently low throughput as only one retina can be live imaged at a time. This bottleneck will be especially evident when time-lapse recordings are conducted for $ 12 h.
In silico trajectory verification presents another potential bottleneck in the protocol as we recommend verifying each individually generated trajectory. Generally, trajectory generation reliability increases with higher signal-to-noise ratios in the xyzt image series. Producing high-quality timelapse recordings as well as appropriate image processing workflows are thus essential to increasing the data output.

TROUBLESHOOTING Problem 1
Step 1.a.: Neural retina prolapse from eyeball during its isolation resulting in retinal tissue damage.

Potential solution
During eyeball dissection to isolate neural retina, the eyeball needs to be carefully opened using dissection tools as described. Small amounts of pressure applied to the eyeball during this step can lead to the neural retina prolapsing from introduced eyeball openings, compromising retinal integrity. This can be avoided by piercing the eyeball at the interface between cornea and sclera using a fine gauge needle (e.g., 30 G BD microlance 3 needle, Cat#304000).

Potential solution
In case xyz tissue drift is detected during time-lapse recording, adjust field of view and/or upper/ lower limits of z-stack accordingly in the intervals between time series points (this may require prematurely stopping the acquisition process and initiating a new xyzt acquisition, which can subsequently be concatenated in silico with the initial acquisition, for instance using Fiji/ImageJ). To avoid tissue drift in future experiments, consider the following options during tissue mounting (steps 2. and 3. in step-by-step method details): Using a heavier platinum harp.
More firmly pressing the edges of the nitrocellulose membrane with mounted retina onto the vaccume grease layer using a round edge of curved forceps.
Securing the edge of nitrocellulose membrane with mounted retina with extra vacuum grease contiguously spread over both the nitrocellulose edge and the underlying petri dish.

Problem 3
Step 11: signs of boiling/bubbling evident during time-lapse recording.

Potential solution
Boiling and bubbling are usually a consequence of pigmented cells (usually black in colour) present within the field of view. These absorb infrared light and rapidly heat up the tissue leading to irreversible tissue damage. To prevent this from occurring, consider imaging a field of view completely devoid of pigmented cells or completely removing all pigmented tissues/cells from neural retina during relevant dissection steps and/or prior to live imaging.

Problem 4
Step 11: loss of retinal integrity during time-lapse recording.

Potential solution
Even the most benign tissue damage introduced during retinal dissection and mounting tends to compound during live imaging, eventually leading to loss of retinal integrity. This can be mitigated by careful tissue handling and/or by avoiding damaged areas of the retina during live imaging.

Problem 5
Step 15: Spot tracking in IMARIS produces too many tracking artifacts, misrepresenting true somal movements.

Potential solution
Using an optimized workflow in Fiji/ImageJ can contribute greatly towards generating higher quality, high signal-to-noise time-lapse recordings post acquisition, which can improve subsequent tracking in IMARIS. While there is no 'one-size-fits-all' approach applicable to all time-lapse experiments, we have found the 'Remove outliers' tool in Fiji/ImageJ particularly useful to remove noise artifacts that can impede faithful tracking in IMARIS. Set parameters to remove ''bright'' outliers at a radius and threshold that approximately match the characteristics of noise artifacts.
Furthermore, the spot detection tool in IMARIS offers filter options to remove artifactual trajectories. The 'Quality' filter can be utilized to retain only high-quality tracking points, rather than lower-quality tracking points, which are more likely to be associated with imaging noise and/or background signal. Other IMARIS filters that proved to be useful for tracking photoreceptor somata that primarily exhibit radial (along the z axis) motility were x and y dimension filters. These can be used to restrict the maximum permissible amount of tangential (xy) motility in the generated tracks, which further eliminated artifactual tracking in our experiments.

RESOURCE AVAILABILITY
Lead contact Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Rachael Pearson (rachael.pearson@kcl.ac.uk).

Materials availability
This study did not generate new unique reagents.

Data and code availability
The data sets generated during and/or analysed during the current study are available from the lead author on request. The code generated during this study is available on the project Github repository on https://github.com/RPearsonLab/Photoreceptor_tracking.

DECLARATION OF INTERESTS
The authors declare no competing interests.