Dendritic localization of mRNA in Drosophila Mushroom Body Output Neurons

Memory-relevant neuronal plasticity is believed to require local translation of new proteins at synapses. Understanding this process requires the visualization of the relevant mRNAs within these neuronal compartments. Here we used single-molecule fluorescence in situ hybridization (smFISH) to localize mRNAs at subcellular resolution in the adult Drosophila brain. mRNAs for subunits of nicotinic acetylcholine receptors and kinases could be detected within the dendrites of co-labelled Mushroom Body Output Neurons (MBONs) and their relative abundance showed cell-specificity. Moreover, aversive olfactory learning produced a transient increase in the level of CaMKII mRNA within the dendritic compartments of the γ5β′2a MBONs. Localization of specific mRNAs in MBONs before and after learning represents a critical step towards deciphering the role of dendritic translation in the neuronal plasticity underlying behavioural change in Drosophila.


Introduction
Memories are believed to be encoded as changes in the efficacy of specific synaptic connections. Dendritic localization of mRNA facilitates specificity of synaptic plasticity by enabling postsynaptic synthesis of new proteins where and when they are required (Holt et al., 2019). Visualizing individual dendritically localized mRNAs in memory-relevant neurons is therefore crucial to understanding this process of neuronal plasticity.
Single molecule fluorescence in situ hybridization (smFISH) enables cellular mRNAs to be imaged at single-molecule resolution through the hybridization of a set of complementary oligonucleotide probes, each labelled with a fluorescent dye. Recent improvements in smFISH permit mRNA transcripts to be visualized in the dense heterogenous tissue of intact Drosophila brains (Long et al., 2017;Yang et al., 2017). Combining whole fly brain smFISH with neuron-specific co-labelling makes Drosophila an ideal model to investigate cell-specific mRNA localization and whether it is regulated in response to experience.
Olfactory learning in Drosophila depresses cholinergic synaptic connections between odorspecific mushroom body Kenyon Cells (KCs) and MBONs (Cohn et al., 2015;Handler et al., 2019;Hige et al., 2015;Owald et al., 2015;Perisse et al., 2016;Séjourné et al., 2011). This plasticity is driven by dopaminergic neurons whose presynaptic terminals innervate anatomically discrete compartments of the mushroom body, where they overlap with the dendrites of particular MBONs (Aso et al., 2010;Burke et al., 2012;Claridge-Chang et al., 2009;Lin et al., 2014;Liu et al., 2012). Dopamine driven plasticity is mediated by cAMPdependent signalling and associated kinases such as Calcium/Calmodulin-dependent protein kinase II (CaMKII) and Protein Kinase A (PKA) (Boto et al., 2014;Handler et al., not been visualized in neurons. We therefore first hybridized CaMKII smFISH probes to whole mount brains and imaged the MB calyx ( Figure 1A, 1B), a recognisable neuropil containing the densely packed dendrites of ~2,000 KCs and their presynaptic inputs from ~350 cholinergic olfactory projection neurons , using a standard spinning disk confocal microscope. To detect and quantify mRNA within the 3D volume of the brain, we developed a FIJI-compatible custom-built image analysis tool that segments smFISH image data and identifies spots within the 3D volume using a probability-based hypothesis test. This enabled detection of mRNAs with a false discovery rate of 0.05. CaMKII smFISH probes labelled 56 േ 5 discrete puncta within each calyx ( Figure 1B, 1C). In comparison, smFISH probes directed to the α 1 nicotinic acetylcholine receptor (nAChR) subunit labelled 33 േ 2 puncta in the calyx ( Figure 1B, 1C). Puncta were diffraction limited and the signal intensity distribution was unimodal ( Figure 1D-DԢ), indicating that they represent single mRNA molecules.

mRNA localization within Mushroom Body Output Neuron Dendrites
Drosophila learning is considered to be implemented as plasticity of cholinergic KC-MBON synapses. To visualize and quantify mRNA specifically within the dendritic field of the γ 5βԢ2a and γ 1pedc>α/β MBONs we expressed a membrane-tethered UAS-myr::SNAP reporter transgene using MBON-specific GAL4 drivers. This permitted simultaneous fluorescent labelling of mRNA with smFISH probes and the MBON using the SNAP Tag ( Figure 1E). To correct for chromatic misalignment (Matsuda et al., 2018) that results from imaging heterogenous tissue at depth we also co-stained brains with the dsDNA-binding dye Vybrant DyeCycle Violet (VDV). VDV dye has a broad emission spectrum so labelled nuclei can be imaged in both the SNAP MBON and smFISH mRNA channels. This triple-labelling approach allowed quantification and correction of any spatial mismatch between MBON and smFISH channels in x,y and z planes, which ensures that smFISH puncta are accurately assigned within the 3D volume of the MBON dendritic field ( Figure 1F).
Using this smFISH approach we detected an average of 32 േ 2 CaMKII mRNAs ( Figure 1G-GԢ) within the dendrites of γ 5βԢ2a MBONs. However, in contrast to the calyx, we did not detect nAChRߙ1 in γ 5βԢ2a MBON dendrites ( Figure 1H-HԢ). This differential localization of the CaMKII and nAChRߙ1 mRNAs within neurons of the mushroom body is indicative of cellspecificity. To probe mRNA localization in MBONs more broadly, we used a single YFP smFISH probe set and a collection of fly strains harboring YFP insertions in endogenous genes (Lowe et al., 2014). We selected YFP insertions in the CaMKII, PKA-R2, and Ten-m genes as test cases and compared the localization of their YFP-tagged mRNAs between γ 5βԢ2a MBON and γ 1pedc>α/β MBON dendrites.
The CaMKII::YFP allele is heterozygous in flies also expressing myr::SNAP in MBONs.
Therefore, YFP smFISH probes detected half the number of CaMKII mRNAs in γ 5βԢ2a MBON dendrites compared to CaMKII-specific probes (Figure 2A-AԢ, 2C). Importantly, YFP probes hybridized to YFP-negative control brains only produced a background signal ( Figure   2B-BԢ) that was statistically distinguishable in brightness from genuine smFISH puncta ( Figure 2D). These data indicate that the YFP probes have specificity and that the YFP insertion does not impede localization of CaMKII mRNA. We detected a similar abundance of CaMKII::YFP in the dendritic field of γ 1pedc>α/β MBONs. In contrast, we detected more PKA-R2 mRNAs in γ 5βԢ2a MBONs compared to γ 1pedc>α/β MBONs. Surprisingly, Ten-m mRNAs were not detected in either γ 5βԢ2a and γ 1pedc>α/β MBON dendrites ( Figure 2E, 2I), although they were visible in neighboring neuropil.
Although we did not detect nAChRߙ1 mRNA within γ 5βԢ2a MBON dendrites, prior work has shown that nAChR subunits, including nAChRα1, are required in γ 5βԢ2a MBON postsynapses to register odor-evoked responses and direct odor-driven behaviours . Since the YFP insertion collection does not include nAChR subunits, we designed nAChRߙ5 and nAChRߙ6 specific smFISH probes. These probes detected nAchRߙ5 and nAchRߙ6 mRNAs within γ 5βԢ2a and γ 1pedc>α/β MBON dendrites, with nAchRߙ6 being most abundant ( Figure 2F, 2I). The selective localization of nAchRߙ5 and nAchRߙ6 mRNA to MBON dendrites indicates that these receptor subunits may be locally translated to modify the subunit composition of postsynaptic nAChR receptors.
Localized mRNAs were on average 2.8x more abundant in γ 5βԢ2a relative to γ 1pedc>α/β MBON dendrites ( Figure 2E, 2F). We therefore tested whether this apparent differential localization correlated with dendritic volume and/or the number of postsynapses between these MBONs. Using the recently published electron microscope volume of the Drosophila Larger dendritic field volume and synapse number is therefore correlated with an increased number of localized mRNAs, suggesting that these parameters may be important determinants of localized mRNA copy number.

Learning transiently changes CAMKII mRNA abundance in 5Ԣ2a MBON dendrites
We tested whether CaMKII::YFP mRNA abundance in γ 5β ᇱ 2a and γ 1pedc>α/β MBONs was altered following aversive learning ( Figure 3A, 3B). We also quantified mRNA in the soma and nuclei of these MBONs ( Figure 3AԢ, 3BԢ). Transcriptional activity is indicated by a bright nuclear transcription focus ( Figure 3C). We subjected flies to four conditions ( Figure 3D): 1) an 'untrained' group that were loaded and removed from the T-maze but not exposed to odors or shock. 2) an 'odor only' group, exposed to the two odors as in training but without shock. 3) a 'shock only' group that were handled as in training and received the shock delivery but no odor exposure. 4) a 'trained' group that were aversively conditioned by pairing one of the two odors with shock. Fly brains were extracted 10 min, 1 h or 2 h after training and processed for smFISH.
CaMKII mRNA increased significantly in γ 5βԢ2a MBON dendrites 10 min after training ( Figure 3E) compared to all control groups. Levels returned to baseline by 1 h and remained at that level 2 h after training ( Figure 3E). CaMKII mRNAs in γ 5βԢ2a MBON soma showed a different temporal dynamic, with transcripts peaking 1 h after training, albeit only relative to untrained and odor only controls ( Figure 3E). The proportion of γ 5βԢ2a nuclei containing a CaMKII transcription focus did not differ between treatments ( Figure 3E), suggesting that the transcript increase in the soma is not correlated with the number of actively transcribing γ 5βԢ2a nuclei, at least at the timepoints measured. Assessing CaMKII mRNA abundance in γ 1pedc>α/β MBONs after learning did not reveal a change in mRNA abundance in the dendrites or soma between trained flies and all control groups at all timepoints measured ( Figure 3F). These results indicate specificity to the response observed in the γ 5βԢ2a MBONs.
Since CaMKII protein is also labelled with YFP in CaMKII::YFP flies, we also assessed protein expression by measuring YFP fluorescence intensity specifically within the MBON dendrites. This analysis did not reveal a significant difference in fluorescence intensity across treatments (Supplementary Figure 1). Since smFISH provides single-molecule estimates of mRNA abundance, a similar level of single-molecule sensitivity may be required to detect subcellular resolution changes in protein copy number.
Prior studies have demonstrated a learning-related increase in the response of γ 5βԢ2a MBONs to the shock-paired odor and suggested this arises from release of feedforward inhibition from γ 1pedc>α/β MBONs (Bouzaiane et al., 2015;Owald et al., 2015). We speculate that the change in CaMKII mRNA abundance might be a consequence of network-level potentiation of the activity of the γ 5βԢ2a MBON, such as that that would result from a release from inhibition.

Acknowledgements
We are grateful for the microscopy facilities and expertise provided by Micron Advanced

Declaration of Interests
The authors declare no competing interests.

Fly Strains
Flies were raised on standard cornmeal agar food in plastic vials at 25 ˚C and 40-50 % relative humidity on a 12 h :12 h light : dark cycle. Details of fly strains are listed in the Key Recourses Table. smFISH probes Oligonucleotide probe sets were designed using the web-based probe design software https://www.biosearchtech.com/stellaris-designer. The YFP smFISH probe set was purchased from LGC BioSearch Technologies (California, USA) prelabelled with Quasar-670 dye. CaMKII, nAChRߙ1, nAChRߙ5 and nAChRߙ6 DNA oligonucleotide sets were synthesised by Sigma-Aldrich (Merck) and enzymatically labelled with ATTO-633 according to (Gaspar et al., 2017). DNA oligonucleotide sequences for each smFISH probe set are provided in the Supplementary Information.

Whole Drosophila brain smFISH
Whole adult brain smFISH was performed essentially as described (Yang et al., 2017

Olfactory conditioning
Aversive olfactory conditioning was performed essentially as described by (Tully & Quinn, 1985

Deconvolution
Deconvolution was carried out using commercially available software (Huygens Professional v19.10.0p1, SVI Delft, The Netherlands). Raw image data generated in .mvd2 file format were converted to OME.tiff format using FIJI (Schindelin et al., 2012) (convert_mvd2_to_tif.ijm). Spherical aberration was estimated from the microscope parameters (see Microscopy). Z-dependent momentum preserving deconvolution (CLME algorithm, theoretical high-NA PSF, iteration optimized with quality change threshold 0.1% and iterations 40 maximum, signal to noise ratio 20, area radius of background estimation is 700 nm, a brick mode is 1 PSF per brick, single array detector with reduction mode SuperXY) was then applied to compensate for the depth-dependent distortion in point spread function thereby reducing artefacts and increasing image sharpness.

Multi-channel alignment
Misalignment between channels was corrected for using Chromagnon (v 0.81) (Matsuda et al., 2018). To estimate channel misalignment nuclei were labelled with the broad emission spectrum dye (Vybrant DyeCycle Violet Stain, ThermoFisher) (Smith et al., 2015). The dye was excited at 405nm and emission was recorded using the appropriate filters for each imaging channel. Chromatic shift was estimated by finding the affine transformation that delivers a minimum mean square error between the nuclear stain in the various channels.
Nuclear calibration channels for chromatic shift correction were separated using ImageJ (see macro Split_ometiff_channels_for_chromcorrect.ijm). The affine transformation was estimated and alignment was performed by calling Chromagnon from Python (see script chromagnon_bash.py). The resulting aligned and deconvolved images were saved in .dv format for further downstream analysis. number of postsynapses within these compartments was also determined using the synapse data that accompany the neuron skeletons (Xu et al., 2020).

mRNA detection
localized mRNA transcripts in Drosophila brains. Software for processing smFISH datasets is available as Supplementary Software. Updates will be made at https:// github.com/qnano/smfish. The smFISH channel was extracted and stored as a 3D greyscale image. mRNA signal was detected using 3D generalized likelihood ratio test (Smith et al., 2015).The false detection rate is 0.05 and the spot width is σ x,y = 1.39 and σ z = 3.48. After 3D detection, the intensity, background, width and subpixel position of the detected mRNA spots are estimated using maximum likelihood estimation (MLE) (Smith et al., 2010).To reduce the impact of overlapping spots in 3D, only a 2D cross-section is used from the zplane where the spot is detected. To filter out spurious detections all spots with a width greater than 5 pixels are discarded.

mRNA-dendrite co-localization
To quantify calyx and dendritic localized smFISH puncta, the calyx and dendritic area were first segmented manually. The contour of the calyx and dendritic area is converted to a mask (M 1 ) using the MATLAB R2019b function roipoly. To quantify smFISH puncta co-localizing with dendrite label, a mask of the dendrite label is created by enhancing the image using a difference of Gaussians filter (width of 1 and 5 pixels) and then thresholding the product between the enhanced image (A) and masked area (M ଵ ) to obtain a mask (M 2 ):

Spot brightness and full width half maximum (FWHM) analysis
For each detection, a region of interest (ROI) is extracted as a 2D box in the x-y plane with a size of 2×(3σ x , y +1). For each ROI the MLE of the x and y position, the number of photons, the number of background photons, and the width of the 2D Gaussian, σ x , y is computed. The FWHM of the spots is calculated as .

Verification of transcription foci
Soma containing bright nuclear transcription foci were selected to quantify the difference in intensity relative to diffraction-limited smFISH puncta. The nuclear localization of the smFISH puncta with the highest photon count was validated by visual inspection and was considered to correspond to the transcription site. The width (σ x,y ) of the transcription foci significantly differ from the sparse smFISH signal and is estimated by fitting a 2D Gaussian to the transcription site using the MATLAB 2019b non-linear least-squares routine lsqcurvefit. Transcription foci brightness and background were computed using the same MLE protocol as for diffraction-limited spots, but with the estimated σ x,y .

YFP fluorescence intensity
To quantify YFP fluorescence intensity within co-labelled neurons, we developed a FIJI compatible macro plugin. Depth-dependent bleaching was first corrected for over the z-stack using an exponential fit. Background signal was then subtracted in each z-section using a rolling ball filter with a width of 60 pixels. 5 z-sections above and below the centre of the image were cropped for analysis. YFP fluorescence intensity was recorded within the dendrites or soma of the co-labelled neuron using the mask described above (mRNAdendrite co-localization). Fluorescence intensity was calculated as analog digital units (adu)/ volume (dendrites or soma) to give adu/voxel. Software for analysing fluorescent protein expression in single neurons is available as Supplementary Software. Updates will be made at https:// github.com/qnano/smfish.

Statistical analyses
Data were visualized and analysed statistically using GraphPad Prism Version 8.3.1 (332).
The distribution of a dataset was assessed with a Shapiro-Wilk test. All smFISH abundance data followed a Gaussian distribution. smFISH abundance was compared between two groups using an unpaired t-test. smFISH abundance between multiple groups was compared using a one-way ANOVA followed by Tukey's post hoc test. Proportions of transcriptionally active soma were compared to transcriptionally inactive soma using a Chi-Square test. YFP positive and negative smFISH intensity distributions were compared with a two-sided Wilcoxon rank-sum test. YFP fluorescence intensity across treatments was compared using a one-way ANOVA for Gaussian distributed data, and a Kruskal-Wallis test for non-Gaussian distributed data. Statistical significance is defined as p<0.05.