A GFP Complementation-based Dual-expression System for Assessing Cell-Cell Contact Mediated by Cytonemes in Live Drosophila Wing Imaginal Discs

The development of embryonic tissues is controlled by cells located in 'organizing centers' that signal to distant cells within a tissue, controlling their decisions to proliferate (i.e., grow and divide) or adopt particular fates¹. This cell non-autonomous signaling is mediated by ligands produced by organizing center cells that form concentration gradients through the tissues and elicit concentration-dependent responses. In many cases, these ligands are either delivered or picked up through long actin-based signaling filopodia called cytonemes that connect signal-sending and -receiving cells in tissues^2,3. First discovered in the Drosophila wing imaginal disc⁴, cytonemes have also been identified in mammals and other vertebrates^5,6,7,8,9. A better understanding of the role of cytonemes in cell non-autonomous signaling, while at an early stage, is crucial to deciphering how cells communicate to organize into tissues and how these communication lines are modified in various pathological conditions, including developmental malformations and cancer.

Cytonemes can extend from source cells to deliver ligands to target cells or from target cells to receive ligands close to their sources^2,3. Cytonemes make intimate contacts with their targets, where they are thought to form synapse-like structures, where ligand transfer can occur^3,10,11. This contact can occur between the tips of source and target cytonemes or between cytonemes and cell bodies³. Although not extensively characterized, cell adhesion through adhesion molecules or through receptor-ligand interactions is, in some cases, needed for the proper activation of downstream signaling events^12,13,14, making this an important aspect of cytoneme biology.

Several studies have applied the "GFP reconstitution across synaptic partners" (GRASP) technique to the analysis of cytoneme contacts. This method was developed for identifying and mapping synaptic partners in complex nervous systems¹⁵. It is based on the expression of the two complementary fragments of split-GFP (spGFP1-10 and spGFP11), each fused to the extracellular region of a transmembrane domain (e.g. of CD4), in different populations of cells. If the plasma membranes of cells in those two populations come into direct contact, it brings the complementary domains of spGFP into proximity, leading to the reconstitution of GFP fluorescence. This approach has been used in Drosophila to identify the existence of cytoneme contacts between cells within the wing disc and between the wing disc and other closely apposed tissues^{12,16,17,18,19,20,21}.

This paper describes the application of GRASP to the characterization of contacts between two morphologically distinct epithelial layers of the Drosophila wing imaginal disc, the disc proper (DP) and the peripodial membrane (PerM). These epithelial layers form a sac surrounding a central lumen, with the pseudostratified columnar DP cells located on one side and the squamous PerM cells on the other, both with their apical membranes facing inwards towards the lumen (Figure 1A). There is some evidence for translumenal signaling between the two layers^22,23,24,25, and we recently documented signaling from the DP to control the proliferation of PerM cells that is mediated by apical cytonemes in the DP²¹. This protocol involves using the GAL4/UAS and LexA/LexAop transgene expression systems to express complementary fragments of spGFP fused to CD4 on the membranes of DP and PerM cells. It uses reconstituted GFP fluorescence to readout membrane contact between the two cell populations.

The nubbin-GAL4 driver is used to express CD4-spGFP1-10 specifically in the wing pouch region of the DP (Figure 1A). The PerM-LexA driver²¹ is used to express CD4-spGFP11 specifically in the PerM (Figure 1A). These two expression systems are independent of one another, allowing simultaneous and specific expression of different transgenes in DP and PerM (Figure 1B,C).

The basic genetic scheme involves crossing flies to generate larvae of the genotype nub-GAL4/UAS-CD4-spGFP1-10;PerM-LexA/LexAop-CD4-spGFP11. As a negative control, we leave out the LexAop-CD4-spGFP11 transgene. Other transgenes (e.g., protein-coding, double-stranded RNA) can be expressed as desired in the DP layer (under the control of UAS sequences) or in the PerM (under the control of the LexA operator).

This is a live imaging protocol that cannot be interrupted. Material preparation is estimated at ~10 min. Dissections should not be performed for more than 20 min at a time before imaging. Imaging takes ~30 min and should not last more than ~1 h. For multiple conditions or a high number of samples, the procedure must be carried out in multiple rounds to ensure the best results.

1. Material preparation

1. Ahead of dissection, clean tweezers, a 9-well glass depression spot plate (or other container suitable for collecting and cleaning larvae), and a reusable Sylgard (silicone)-coated Petri dish used for dissection with 70% ethanol.
2. Prepare microscope slides by performing a 70% ethanol wash and then sticking an imaging spacer to the dried slide. Cut individual wells out of the 8-well strip with scissors and cut in half (Figure 2A). Remove the protective backing from one side of each half and stick spacers to the slide, spaced slightly apart to create a sheltered space in the middle where the discs will be placed (Figure 2B).
   NOTE: In our experience, this simplifies sample mounting as the precise volume of medium transferred to the slide is less critical. Any small excess volume will simply flow out through the gap upon coverslipping. As we image for only a relatively short time (generally ~1 h), we did not experience any problems with medium drying.
3. Thaw live-imaging medium (unsupplemented Schneider's Drosophila Medium stored in frozen 10 mL aliquots) and keep it readily accessible in an ice bucket.
4. In a 2 mL microcentrifuge tube, mix 1.5 mL of medium with 20 µL of 200 µg/mL Hoechst 33342 (final concentration 2.7 µg/mL). Keep the tube on ice.
   NOTE: This specific form of Hoechst dye is used because it is highly membrane-permeable. CAUTION: Gloves should be used when handling Hoechst 33342. Tools should be washed after use.

2. Wing disc dissection and slide preparation

1. Following genetic crossing and rearing at 25°C, pick wandering third-instar larvae from the tube and wash them in cold live-imaging medium in a 9-well glass depression spot plate.
   NOTE: Cytonemes are very fragile. In the following steps, care should be taken to manipulate tissues around the discs, trying to avoid direct contact between the discs and dissection tools. We prepare three or four larvae at a time, as damaging a few discs is unavoidable, and this may only become noticeable during the imaging steps.
2. Under the dissecting microscope, transfer the larvae into a drop of live-imaging medium on a silicone-coated Petri dish and start dissecting using two dissecting forceps. To do so, first hold the head and anterior body part steady by pinching the larvae with one pair of forceps at about one-third the body length (from the anterior). Using the second pair of forceps, pinch the body just posterior to the first forceps. Then, pull the posterior part of the larvae away to isolate the anterior "half".
3. Invert the anterior half by securing it with one pair of tweezers on either side of the cut end and pushing the head through with the other pair (like turning a sock inside out), to expose the internal structures, including imaginal discs, trachea, salivary glands, fat body, and gut.
4. Carefully remove the salivary glands, fat body, and gut, being careful not to disturb the lateral trunks of the trachea, which should overlie and protect the wing discs.
5. Use forceps to carefully transfer the cleaned anterior halves with attached wing discs into a drop of clean live-imaging medium with Hoechst, placed between the spacers on a prepared slide (Figure 2C). Isolate the wing discs from the rest of the tissues by gently blunt-dissecting them away from the fine branches of the trachea that connect to the disc, using either one blade of the dissecting forceps or a fine tungsten wire attached to a dissecting needle holder. Discard the rest of the carcass.
6. Orient the discs using one blade of the dissecting forceps or tungsten wire dissecting needle. Due to their flattened tear-drop-like shape, wing discs will fall with either the DP or PerM facing up. For this protocol, orient all discs with the PerM side up.
7. Adjust the medium volume to fill the well entirely, to slightly above the level of the imaging spacer, by either adding or removing the medium as needed. Remove the protective backing from the top side of each imaging spacer half and then carefully lower a coverslip over the sample. Gently press on the coverslip where it contacts the imaging spacers (e.g., with the rounded end of dissecting forceps) to ensure that the coverslip adheres to the spacer.
   NOTE: The medium level should be slightly above the spacer height to avoid trapping air bubbles, but not so high as to generate significant flow when the coverslip is lowered.

3. Imaging

1. Image samples at room temperature using a confocal microscope (argon ion laser with a 488 nm wavelength for GFP, UV laser with 350 nm wavelength for Hoechst) equipped with a 40x/1.30 NA oil immersion lens and microscopy software. Acquire image stacks with 1 µm step size that span from just above the highest level of the PerM (where Hoechst fluorescence diminishes) to the basal side of the DP, which generally works out to more than 100 images per stack. Acquire repeated image stacks to generate timelapse movies.
   NOTE: We image a slide for no more than ~1-2 h.
2. For the microscope software settings, acquire images using 8-bit coding, line averaging of 2, detector gain generally between 650 and 750, and separated channel tracks to reduce signal bleedthrough. Use identical laser and detector settings for all conditions, predetermined by performing tests on each condition beforehand.
   NOTE: At 40x magnification, our images have a resolution of 5.0386 pixels per µm (pixel size: 0.1985 x 0.1985 µm²). In our setup, the optimal confocal microscope settings included the highest laser intensity possible (without damaging the tissue) to capture as much signal as possible, while at the same time yielding little to no saturation; and the lowest possible noise as assessed on negative controls, achieved by reducing detector offset and acquisition speed (if needed). We generally use less than 5% laser power and maximum scan head speed ('9' in the Zen software settings). A single image stack typically takes ~1 min 45 s to acquire, and images are taken every 2-3 min for timelapse analyses.

4. Image analysis

1. Project stacks using the maximum intensity projection function of the ImageJ software package (click on image | Stacks | Z project, select Maximum intensity in the menu).
2. In the MIP images, determine the wing pouch area based on the pattern of wing disc folding (using the Hoechst channel) using Polygon Selection. Use the same selection on the GFP channel with the Clear Outside Function (click on Edit | Clear outside) to set every outside pixel value to 0.
3. In the MIP images, use the Hoechst channel to clean the GFP images, specifically removing any artefactual spots that appear in both channels by setting the concerned pixel value to 0 (click on Polygon Selection, then Edit | clear).
4. In the cleaned MIP images, use the Histogram function (click on Analyze | Histogram | List) to obtain and copy the list of all pixel values into a table in a spreadsheet.
5. Set a threshold value based on the analysis of the background signal in negative controls (and then verified in other samples; see the discussion).
6. Calculate the percentage of GFP-positive surface using the following formula:
   %GFP-positive surface = × 100
7. Normalize the data at the end by dividing each value by the mean of the reference condition (which is set to 1).
   NOTE: Valid pixel = pixel with a value ≥ 1, effectively ignoring any deleted pixel whose value was set to 0. Pixel values of 0 do not occur naturally.
8. For determining where in the discs the reconstituted GFP signal localizes, use XZ-reslices of image stacks to look at the apicobasal axis of the disc epithelium. Perform the reslices in Fiji without interpolation, and increase the image height in the Z-axis using the Scale function with bilinear interpolation to improve visibility.

To test the usefulness of the GRASP procedure for measuring contacts between DP and PerM cells, we examined wing discs of four different genotypes: wild-type negative-control discs (genotype: w1118) which will only display background levels of autofluorescence in the GFP channel; discs expressing the CD4-spGFP1-10 in the DP layer, but lacking the CD4-spGFP11 transgene, which will show the level of fluorescence produced by GFP1-10 alone (which we expected to be negligible, as GFP1-10 should not fluoresce26); discs in which CD4-spGFP1-10 and CD4-spGFP11 are expressed in DP and PerM, respectively (Figure 3A), which will reveal the "normal" level of DP-PerM contact; and discs which, in addition to CD4-spGFP1-10 and CD4-spGFP11, also express the Ser/Thr protein kinase Slik in DP cells under UAS control. Slik expression in DP cells drives non-autonomous proliferation of PerM cells27. It also drives the formation of apical cytonemes in DP cells that cross the disc lumen and appear to make contact with PerM cells, often stably21. We thus expected Slik expression to enhance GFP complementation in this experimental setup.

Imaging at the level of the PerM in wild-type discs revealed the usual arrangement of these cells, whose squamous morphology and low proliferation rate result in a spread-out appearance of nuclei (visualized by Hoechst staining, Figure 3B). There was a minimal amount of autofluorescence detected, mainly in the form of granular spots of low intensity that tended to be clustered towards the middle of the wing pouch (Figure 3B). Discs expressing only CD4-spGFP1-10 in the DP looked very similar, and the fluorescence appeared to be of a similar intensity (Figure 3C). We used the intensity level of the autofluorescence in negative control discs to set a threshold for signals in the experiment, with less than ~0.2% of pixel intensities in most negative-control images falling above this threshold. Quantification of the relative area of the wing pouch region in each image that had pixel values above this threshold confirmed that the level of fluorescence was indistinguishable from the background in discs expressing only CD4-spGFP1-10 (Figure 3F).

In discs expressing CD4-spGFP1-10 in the wing pouch and CD4-spGFP11 in the PerM, we saw the appearance of more and brighter spots of GFP fluorescence (Figure 3D). In XZ-reslices of disc image stacks, these brighter spots localized to the disc lumen, where we expect the GFP complementation to occur21. Quantification revealed that the relative area of the wing pouch region in each image that had pixel values above the threshold was generally greater than 0.5%, a level that was significantly higher than the negative controls (Figure 3F). This result suggests that there is normally a low level of contact between DP and PerM cells during disc development.

Expression of Slik in the DP layer caused dramatic changes to the appearance of the PerM. In the region overlying the wing pouch where Slik was expressed, there was a huge increase in the number of PerM cell nuclei, reflective of a strong stimulation of cell proliferation (Figure 3E). There was also a widespread appearance of high-intensity GFP fluorescence that matched the location of increased PerM cell proliferation (Figure 3E,F). This suggests that the apical cytonemes formed in response to Slik do make close contacts with cells in the PerM, consistent with the proliferative response of these cells. While it is clear from these results that Slik promotes DP-PerM membrane-membrane contacts, we cannot rule out that other mechanisms, such as secreted or shed DP membrane vesicles, may contribute to the effect.

Figure 1: Organization of the Drosophila wing imaginal disc. (A) Schematic of a wing imaginal disc. Image on the left is an en face (XY) view, on the right is a YZ cross-section. The DP is in blue, with the wing pouch region shaded darker blue. The PerM is represented by a black dotted line in the XZ view. The DP and PerM enclose a central lumen. (B) MAX Z-projection of a confocal image stack of a wing disc in which myristoylated TdTomato (red) was expressed throughout the wing pouch under the control of the nubbin-GAL4 driver, and nuclear GFP (GFPnls, green) was expressed in the PerM under the control of the PerM-LexA driver. Scale bar = 20 µm. (C) XZ reslice of a disc image stack as in B, showing non-overlap of the nubbin-GAL4 and PerM-LexA expression domains. Scale bar = 20 µm. Please click here to view a larger version of this figure.

Figure 2: Mounting of samples with spacers for live imaging. Photos showing (A) cutting of the 8-well imaging spacers, (B) positioning and affixing of the spacers on a glass microscope slide, and (C) positioning of the drop of live-imaging medium within the space sheltered by the imaging spacers, where discs are mounted. Please click here to view a larger version of this figure.

Figure 3: Quantification of membrane contacts between DP and PerM cell populations in the wing disc. (A) Schematic of experimental setup, with GAL4-driven expression of CD4-spGFP1-10 in the DP and LexA-driven expression of CD4-spGFP11 in the PerM. Contact between the two epithelial layers across the lumen results in GFP fluorescence complementation. (B-E) Confocal images showing merged channels of live wing discs expressing (B) neither portion of spGFP (w1118), (C) CD4-spGFP1-10 in the DP layer alone, (D) together with CD4-spGFP11 in the PerM, or (E) with both CD4-spGFP11 in the PerM and Slik in the DP. Images are MAX Z-projections, with Hoechst in white and GFP channel in green in the top images. Scale bars = 20 µm. (F) Graph representing the relative area of GFP fluorescence above threshold intensity per disc for each condition (mean ± standard deviation). Data were normalized to the CD4-spGFP1-10 + CD4-spGFP11 (baseline) condition. p-values were calculated using unpaired one-way Welch ANOVA with Dunnett T3 multiple comparison. Abbreviations: DP = disc proper; PerM = peripodial membrane. This figure was modified from Rambaud et al.21. Please click here to view a larger version of this figure.

Cytonemes play an important role in the distribution of ligands, controlling the growth and organization of developing tissues. Signal exchange takes place where cytoneme tips make intimate membrane contacts with their targets. In this protocol, we describe a simple method for analyzing cytoneme-mediated contacts between epithelial layers in the wing disc using the GRASP technique.

The technique presented here requires, at a minimum, four components-a GAL4 driver, a LexA driver, and the two transgenes encoding CD4 fusions of the complementary spGFP fragments under UAS and LexAop control. We did not detect any discernible effect or lethality associated with the expression of either of the CD4-spGFP proteins using the drivers we selected. It is thus possible to construct a universal driver stock stably containing all four components. This stock could be crossed to stocks bearing any combination of additional UAS and LexAop transgenes to manipulate gene expression in DP and/or PerM cells to test the effects on cytoneme contact formation. In the process of stock-building (especially in making recombinant chromosomes carrying two components), it may be desirable to detect the expression of just one of the complementary spGFP fragments. Our anti-GFP antibodies did not seem to detect spGFP1-10 protein in paraformaldehyde-fixed discs. However, antibodies that do are available. Monoclonal antibody GFP-G1 from the Developmental Studies Hybridoma bank recognizes spGFP1-10. Antibodies that recognize the GFP11 polypeptide (which is only 16 amino acids long) have also been developed (e.g., PA5-109258 from Invitrogen).

We note that the levels of GFP fluorescence observed in these experiments are very low compared to simply overexpressing GFP with the same drivers. This is because the reconstituted GFP signal was limited to a very restricted location in the disc, mainly at the interface between DP and PerM cells in the disc lumen. GFP fluorescence in the head region of larvae (where wing discs reside) was just detectable in a fluorescence dissecting microscope in discs expressing spGFP1-10 and spGFP11, and was clear, if faint, in discs when Slik was expressed. It is advisable when building stocks to include appropriate balancers with larval markers, fluorescent or otherwise, to permit accurate genotyping of larvae prior to dissection.

Some optimization is needed at each phase of the procedure (sample preparation, imaging, image analysis). As the protocol involves imaging with live samples, and cytonemes are very fragile, much care is needed when dissecting to avoid handling the wing discs themselves. Working in small batches to keep dissecting, mounting, and imaging times to a minimum is critical, with a full cycle generally lasting only 1-1.5 h. For sample mounting, we tested imaging spacers of different thicknesses. The 0.12 mm thickness we selected was best for mature third instar wing discs, as this did not lead to any compression of the discs but was also not so thick as to allow the samples to drift freely in the medium. However, if imaging smaller or larger discs, either from different time points or from genotypes where disc size is affected, it may be necessary to use imaging spacers of a different thickness.

Finding the optimal imaging settings will also require some investigation. As a specific signal can be very low, higher laser power settings and increased detector gain may be needed, which will produce higher noise in the images. However, if one of the experimental conditions has a much higher signal, as was the case in the experiment presented in Figure 3, those same parameters can lead to saturation, which will not allow a proper estimate of the real reconstituted GFP signal. Pilot experiments with the different experimental samples are recommended to establish the optimal parameters that allow for the best signal without saturation in all conditions, before starting to collect images for analysis.

Once high-quality images are obtained, some analysis is needed to determine an optimal cutoff threshold for the GFP signal. Wild-type discs or discs expressing spGFP1-10 alone can be used for this purpose, as any signal observed in these discs comes from autofluorescence. We determined the range of this background noise by plotting a histogram of all pixel intensity values in these negative control discs and fixing a threshold where the curve dropped toward 0. We chose a value that left roughly 99.8% of pixels in negative control discs below this threshold. Only pixels of intensity above this threshold were considered as 'real' signal.

While this protocol deals specifically with cytoneme contacts between DP and PerM cells in the wing disc, the GRASP technique can be applied to any system where cytonemes are involved. Indeed, this approach has been successfully used in Drosophila to demonstrate cytoneme-mediated contacts between pairs or populations of DP cells in discs^{12,16,17,18,19,20,21}. All that is required is appropriate GAL4 and LexA drivers that express in the two cell populations under study. The experiment we presented involves live imaging of discs at a single time point. However, time-lapse fluorescence imaging could be conducted to study cytoneme contact dynamics in tissues. Although we have performed time-lapse imaging on wing discs with a confocal microscope for up to 2 h, this has mostly been at relatively low time resolution (image every 2 min) due to the time required for acquiring detailed image stacks. This could be improved by restricting the depth of tissue imaged to the space within the lumen where contacts appear to occur, thereby allowing faster and more frequent image acquisition. This could be extended to photobleaching and recovery setups (FRAP) to look at contact stability and turnover. This approach thus has the potential to yield important insights into cytoneme function.