Intact Drosophila central nervous system cellular quantitation reveals sexual dimorphism

  1. Wei Jiao
  2. Gard Spreemann
  3. Evelyne Ruchti
  4. Soumya Banerjee
  5. Samuel Vernon
  6. Ying Shi
  7. R Steven Stowers
  8. Kathryn Hess
  9. Brian D McCabe  Is a corresponding author
  1. Brain Mind Institute, EPFL - Swiss Federal Institute of Technology, Switzerland
  2. Department of Microbiology and Cell Biology, Montana State University, United States

Abstract

Establishing with precision the quantity and identity of the cell types of the brain is a prerequisite for a detailed compendium of gene and protein expression in the central nervous system (CNS). Currently, however, strict quantitation of cell numbers has been achieved only for the nervous system of Caenorhabditis elegans. Here, we describe the development of a synergistic pipeline of molecular genetic, imaging, and computational technologies designed to allow high-throughput, precise quantitation with cellular resolution of reporters of gene expression in intact whole tissues with complex cellular constitutions such as the brain. We have deployed the approach to determine with exactitude the number of functional neurons and glia in the entire intact larval Drosophila CNS, revealing fewer neurons and more glial cells than previously predicted. We also discover an unexpected divergence between the sexes at this juvenile developmental stage, with the female CNS having significantly more neurons than that of males. Topological analysis of our data establishes that this sexual dimorphism extends to deeper features of CNS organisation. We additionally extended our analysis to quantitate the expression of voltage-gated potassium channel family genes throughout the CNS and uncover substantial differences in abundance. Our methodology enables robust and accurate quantification of the number and positioning of cells within intact organs, facilitating sophisticated analysis of cellular identity, diversity, and gene expression characteristics.

Editor's evaluation

This manuscript describes a pipeline involving whole brain imaging, automated neuronal segmentation, and topographical analysis, to assess the number of specific cell types in the larval Drosophila brain. The authors uncover unexpected sexual dimorphism at this early stage. This paper will be of interest to neuroscientists – from those who use larval Drosophila as their study model to others who are generally interested in connectomics and transcriptomics.

https://doi.org/10.7554/eLife.74968.sa0

Introduction

Establishing the precise number of cells in the brain is essential to create organ-wide catalogues of cell types and their gene expression (Lent et al., 2012; Devor et al., 2013). However, apart from the nervous system of the nematode Caenorhabditis elegans (302 neurons, 56 glia) (White et al., 1986), the exact numbers of cells within the central nervous system (CNS) of model organisms or that of humans is currently unknown, with estimates, including those based on extrapolation from direct quantification of brain sub-regions, varying widely (Silbereis et al., 2016; Keller et al., 2018; von Bartheld et al., 2016).

Studies of the CNS of Drosophila melanogaster, which in scale and behavioural repertoire has been viewed as intermediate between nematodes and rodents (Bellen et al., 2010; Alivisatos et al., 2012), currently include large-scale efforts to establish both a neuronal connectome and a cell atlas (Scheffer and Meinertzhagen, 2019; Allen et al., 2020; Li et al., 2022). Nonetheless, the precise number of cells (neurons or glia) in either the smaller larval or larger adult Drosophila CNS, comprised of both a brain and ventral nerve cord (VNC), remain unknown, though several approximations have been suggested. For the larval CNS, a range of 10,000–15,000 active neurons has been proposed (Scott et al., 2001; Meinertzhagen, 2018; Eschbach and Zlatic, 2020) across developmental time points. For adult Drosophila, approximations have been suggested in the range of 100,000–199,000 neurons in the brain (Simpson, 2009; Chiang et al., 2011; Kaiser, 2015; Scheffer and Meinertzhagen, 2019; Raji and Potter, 2021) together with a range of 10,000–20,000 cells in the VNC (Birkholz et al., 2015; Lacin et al., 2019; Bates et al., 2019; Allen et al., 2020). The other major CNS cell type, glia, has been estimated to be approximately 10% of the number of neurons (Kremer et al., 2017; Meinertzhagen, 2018; Raji and Potter, 2021). Given the large diversity of these estimates, precise quantification of the numbers of Drosophila neurons and glia would seem a desirable goal, beginning with the smaller larval CNS, which enables the wide compendium of larval Drosophila behaviours (Gerber et al., 2009; Neckameyer and Bhatt, 2016; Eschbach and Zlatic, 2020; Louis, 2020; Gowda et al., 2021).

Complicating the aspiration to quantitate the Drosophila larval CNS, in addition to the general problem of separating and quantifying primary cell types such as neurons and glia, are two specific confounding factors that limit simple total cell quantification approaches. First, encompassed within and surrounding the larval CNS are dividing neuroblasts, which will give rise to adult neurons (Doe, 2017). Relatedly, imbedded within the larval CNS are substantial numbers of immature adult neurons, observed from electron micrograph reconstructions as having few or no dendrites and axons that terminate in filopodia lacking synapses (Eichler et al., 2017). These immature neurons are unlikely to contribute to larval CNS function and are generally excluded when considering larval neuronal circuit architecture (Eichler et al., 2017; Scheffer and Meinertzhagen, 2019). It has been suggested that only a small fraction of the total number of larval CNS cells may actually contribute to CNS function (Ravenscroft et al., 2020).

Here, we have sought to develop a synergistic molecular genetic, imaging, and computational pipeline designed de novo to allow automated neuron, glia, or other gene expression features to be precisely quantitated with cellular resolution in an intact whole CNS. Central to the approach are high signal-to-noise gene expression reporters that produce a punctate, nucleus-localised output, facilitating downstream automated computational measurements and analysis. Exploiting multiple genetic reagents designed to selectively identify only functional neurons with active synaptic protein expression, we identify substantially fewer neurons than most previous estimates in the Drosophila larval CNS and, in addition, substantially more glia. We also discover a previously unsuspected sexual dimorphism in the numbers of both cell types at larval stages. The generation of whole CNS point clouds from our data enables us to apply the tools of topological data analysis (TDA) to summarise the CNS in terms of multiscale topological structures. Utilisation of these topological summaries in a support vector machine also supports that sexual dimorphism extends to deeper features of CNS organisation. Finally, we applied our pipeline to quantitate the whole CNS expression frequency of the Drosophila family of voltage-gated potassium channels, which revealed divergent channel expression frequencies throughout the CNS. We envision that our method can be employed to allow precise quantitation of gene expression characteristics of the constituent cells of the brain, and potentially other intact whole organs, in a format suitable for sophisticated downstream analysis.

Results

Genetic and imaging tools designed to facilitate automated whole CNS cellular quantitation

To establish a robust quantitative method to measure gene expression frequency and quantify the number of cells that contribute to Drosophila larval CNS function, we sought to develop a pipeline utilising genetic reporters designed to expediate automated neuron and glia quantitation from three-dimensional intact organ images. While membrane-associated reporters are generally employed to label Drosophila neurons (Pfeiffer et al., 2008; Jenett et al., 2012; Ravenscroft et al., 2020), we generated UAS-driven (Brand and Perrimon, 1993; Wang et al., 2012) fluorescent reporters fused to histone proteins (Sherer et al., 2020) to target fluorescence only to the nucleus, in order to facilitate subsequent automated segmentation and counting. Through empirical selection of transgene genomic integration sites, we established a set of reporter lines that produced a strong and specific punctate nucleus signal when expression was induced, with little to no unwanted background expression. We then developed a procedure to capture the entire microdissected larval CNS volume by light sheet microscopy at multiple angles and with high resolution, imaging only animals within the ~2 hr developmental time window of the wandering third instar larval stage (Ainsley et al., 2008). These multiview datasets were then processed to register, fuse, and deconvolve the entire larval CNS volume. The volume was then segmented and cell numbers were automatically quantified (Figure 1a–d).

Intact whole CNS quantitation pipeline schematic and validation.

(a–d) Illustration of an intact whole central nervous system (CNS) genetic, imaging and computational pipeline. (a) Genetic reagents: GAL4 is introduced into the exons of genes encoding synaptic proteins (e.g. bruchpilot [brp]) to capture their expression pattern with high fidelity. GAL4 expression regulates the production of UAS fluorescent-histone reporters, which target to the nucleus of cells, producing a punctate signal. (b) Imaging: the intact CNS is imaged at high resolution using light-sheet microscopy. Images are captured at five different angles at 72° intervals. (c) Assembly: multiview light sheet images are registered, fused, and deconvolved. (d) Quantitation: the volume is segmented, and the nucleus number and relative position are measured. Three-dimensional coordinates of the geometric centre of every nucleus can be calculated to produce a point cloud of nuclei positions.(e–h) Pipeline validation. Three-dimensional images before segmentation (above) and subsequent to segmentation (below) of (e) dopaminergic (TH-GAL4) neurons, (f) serotonergic neurons (Trh-GAL4) and (g) dopa decarboxylase expressing (Ddc-GAL4) neurons. (h) Manual or automated quantification of nuclei numbers in these volumes are similar. Scale squares in (e) and (g) are 100 μm and in (f) is 50 μm. (h) Bars indicate minimum and maximum values.

To evaluate the reliability of the procedure, we began by comparing automated counts of distinct neuronal subtypes with manual counting. We separately labelled all dopaminergic neurons (Figure 1e, TH-GAL4, Video 1; Friggi-Grelin et al., 2003a; Mao and Davis, 2009), serotonergic neurons (Figure 1f, Trh-GAL4, Video 2; Alekseyenko et al., 2010) and neurons that produce both types of neurotransmitter (Figure 1g, Ddc-GAL4, Video 3; Lundell and Hirsh, 1994) in the larval CNS. Quantification revealed a high level of concordance (Figure 1h, +/-0.21%, n=5 for TH-GAL4, +/-1%, n=5 for Trh-GAL4, +/-0.38%, n=6 for Ddc-GAL4, Figure 1—source data 1) between automated and manual measurements of these neuronal subtypes establishing confidence in the procedure.

Video 1
Larval CNS labelled with TH-GAL4.
Video 2
Larval CNS labelled with Trh-GAL4.
Video 3
Larval CNS labelled with Ddc-GAL4.

Number of neurons and glia in the female larval CNS

Encouraged by our neuronal subset quantitation results, we next sought to generate GAL4 lines for genes likely to be expressed only in active larval neurons with synaptic connections but not by neuroblasts or by immature neurons (Figure 2—figure supplement 1). We biased towards generating GAL4 insertions within endogenous genomic loci in order to reproduce endogenous patterns of gene expression with high fidelity.

Bruchpilot (Brp) is a critical presynaptic active zone component widely used to label Drosophila synapses, including for large-scale circuit analyses (Wagh et al., 2006). We employed CRISPR/Cas9 genome editing to insert GAL4 within exon 2 of the brp gene, utilising a T2A self-cleaving peptide sequence (Diao et al., 2015) to efficiently release GAL4. While this exonic insertion generated a hypomorphic allele of brp (data not shown) when homozygous, the line was employed in heterozygotes to capture Brp protein expression with high fidelity. To complement this line, we used the Trojan/MiMIC technique (Diao et al., 2015), to generate a GAL4 insertion in the Syt1 gene, which encodes Synaptotagmin 1 (Littleton et al., 1994), the fast calcium sensor for synaptic neurotransmitter release (Quiñones-Frías and Littleton, 2021). Lastly, we examined a transgenic line where an enhancer of nSyb (neuronal Synaptobrevin) (Deitcher et al., 1998), which encodes an essential presynaptic vSNARE (Südhof and Rothman, 2009), is used to control GAL4 expression (Aso et al., 2014). All three lines were expressed in a similar pattern, labelling a substantial fraction but not all of the total cells in the larval CNS (Figure 2a–c, brp-GAL4 female Video 4, Syt1-GAL4 female Video 5, nSyb-GAL4 female Video 6). These lines contrasted with the widely used elav-GAL4 (Lin and Goodman, 1994), which was expressed in larval neurons, but also apparently in some immature neurons and potentially in some glia as well (Berger et al., 2007; Figure 2—figure supplement 1). To characterise our lines, we examined their expression throughout development, beginning with embryogenesis. We detected no expression from any of the three lines prior to embryonic stage 16 (Figure 2—figure supplement 2a). However, beginning at stage 17 of embryonic development, when synaptic activity begins (Baines and Bate, 1998), all three lines displayed expression in both the CNS and peripheral nervous system (Figure 2—figure supplement 2a). We also examined if these lines were expressed in neural stem cells during larval stages by co-labelling the larval CNS with the transcription factor Deadpan, a neuroblast marker (Bier et al., 1992). We found that labelling by all three lines did not overlap with Deadpan expression (Figure 2—figure supplement 2b), suggesting these lines are not expressed in neuroblasts. We also examined expression of all three lines in the adult brain and, as in the larval CNS, observed labelling of a large fraction but not all of the total cells in the adult brain (Figure 2—figure supplement 2c). Lastly, to ensure that the cells labelled by our lines were exclusively neurons, we compared their expression to that of glial cells labelled by glial specific transcription factor Repo (Xiong et al., 1994; Lin and Potter, 2016) using independent and mutually exclusive QF2 dependent labelling. We found complete exclusion of cells labelled by brp, Syt1 and nSyb GAL4 lines from cells labelled by repo (Figure 2d, brp-GAL4 & repo-QF2 Video 7, Syt1-GAL4 & repo-QF2 Video 8, nSyb-GAL4 & repo-QF2 Video 9), consistent with the brp, Syt1 and nSyb GAL4 lines labelling only neurons that express synaptic protein genes and not glial cells.

Figure 2 with 2 supplements see all
Quantitation of neurons and glia in the whole female larval CNS.

(a–d) Multiview deconvolved images (left) and z-stack projections (right) (colours represent z position) of the central nervous system (CNS) of (a) brp-GAL4, (b) Syt1-GAL4, (c)nSyb-GAL4, and (d) repo-GAL4. (e) Distribution of inter-nuclei distances for each line. (f) Quantification of the number of labelled nuclei in each line. (a–d) left; scale squares (a) and (c) = 50 μm, (b) and (d) = 100 μm; right images identical magnification, scale bar = 100 μm. (f) Bars indicate minimum and maximum values.

Video 4
Female CNS labelled with brp-GAL4.
Video 5
Female CNS labelled with Syt1-GAL4.
Video 6
Female CNS labelled with nSyb-GAL4.
Video 7
Larval CNS labelled with brp-GAL4 and repo-QF2.
Video 8
Larval CNS labelled with Syt1-GAL4 and repo-QF2.
Video 9
Larval CNS labelled with nSyb-GAL4 and repo-QF2.

To further compare these lines, beginning with the CNS of female animals, we calculated three-dimensional coordinates for the geometric centre of all nuclei labelled in the brp, Syt1 and nSyb GAL4 lines to generate point cloud mathematical objects and compared them to point clouds of glial nuclei labelled by the repo-GAL4 line. We then plotted and compared the distributions of inter-nuclei distances in these lines. Using this measurement, we found that the inter-nuclei distance of glial cell nuclei exhibited a unimodal distribution (Figure 2e). In contrast, all three neuronal lines exhibited a bimodal distribution of inter-nuclei distances (Figure 2e). We thus observed two patterns of labelled nuclei, one shared among neuronal lines and the other distinct for glia (Figure 2e), again consistent with these lines labelling different cell types.

We next counted the number of nuclei labelled by these neuronal and glial lines, again beginning with females (Figure 2f). We found that the CNS labelled by brp-GAL4 had 10,776 (±2.65%, n=6) neurons, Syt1-GAL4 had 10,097 (±5.96%, n=5) neurons, and nSyb-GAL4 had 9971 (±1.35%, n=5) neurons (Figure 2f). We tested the statistical difference in the numbers of neurons labelled by these lines and found that while nSyb-GAL4 and Syt1-GAL4 were not statistically different from each other, brp-GAL4 did label significantly more neurons than either Syt1 or nSyb GAL4 lines (brp-GAL4 vs Syt1-GAL4+6.72%, p=0.03, brp-GAL4 vs nSyb-GAL4 +8.07%, p=0.01). Averaging across the lines, we found that the female third instar larval CNS had 10,312 ±5.03%, n=16, neurons (Figure 2—source data 1). To ensure that our method did not introduce bias in dense datasets, we also manually counted a brp-GAL4 labelled CNS and compared it to the automated count. Similar to our experiments with sparse neuronal labelling, we found good agreement between manual and automated quantification with a difference of just 14 neurons (9430 nuclei manual vs 9444 nuclei automated for this individual CNS).

We next counted the number of glia labelled by the repo-GAL4 line (Figure 2d and f, repo-GAL4 female, Video 10). We measured 3860 ±3.37%, n=7, glia in the female CNS (Figure 2—source data 1). This amounted to 37% of the number of neurons, far more than previously estimated (Meinertzhagen, 2018; Raji and Potter, 2021). In sum, we found that the female Drosophila larval CNS had 10,312 neurons, fewer than most previous predictions, and several fold more glia than previously thought.

Video 10
Female CNS labelled with repo-GAL4.

Males have fewer neurons and more glia than females

We next carried out a similar analysis on the CNS of male larvae (Figure 3a–c). We found that brp-GAL4 labelled 9888 (±3.15%, n=5) neurons, Syt1-GAL4 labelled 9012 (±3.8%, n=5) neurons, and nSyb-GAL4 labelled 9286 (±5.38%, n=5) neurons in male larvae (Figure 3e, Figure 3—source data 1). In males, brp-GAL4 did not label significantly more neurons than nSyb-GAL4 but did label more than Syt1-GAL4 (brp-GAL4 vs Syt1-GAL4 +9.72%, p=0.01), while the number of neurons labelled by nSyb-GAL4 was not significantly different from Syt1-GAL4, similar to what we had found in females. Averaging across the lines, we found that the male third instar larval CNS had 9396 ±5.59%, n=15 neurons, significantly fewer than those of females (–9.75%, p<0.0001). This difference was also consistent within individual genotypes with brp-GAL4 labelling (–8.98%, p=0.0008), Syt1-GAL4 labelling (–12.04%, p=0.008) and nSyb-GAL4 labelling (–7.38%, p=0.0182) less neurons in males than in females.

Quantitation of neurons and glia in the male larval CNS and topological comparison of sex differences.

(a–d) Example z-stack projections (colours represent z position) of male larval central nervous system (CNS) of (a) brp-GAL4, (b) Syt1-GAL4, (c) nSyb-GAL4, and (d) repo-GAL4. (e) Quantification of the number of labelled nuclei in each line. (f) The distribution of correlations between the ground truth and the prediction made by the support vector machine (SVM) using topological features is indicative of sexual dimorphism of the higher order structure of neuron point clouds (g) Simpler point cloud features such as properties of the distributions of inter-nuclei distances are not indicative of this. (a–d): identical magnification, scale bar = 100 μm. (e) Bars indicate minimum and maximum values.

We also counted the number of glia labelled by repo-GAL4 in males (Figure 3d and e). We found that males had 4015 ±1.98%, n=6, glia far more than previous estimates (Figure 3—source data 1). The number of glia in the male larval CNS was significantly more than in females (+3.86%, p=0.0284). In summary, male Drosophila larva have significantly fewer CNS neurons than females but more glia.

Topological analysis detects CNS structural differences between males and females

We next wished to determine if the differences between point clouds derived from the positions of neuronal nuclei of the male and female CNS went beyond simple numerics. To do this, we applied the tools of TDA (Rabadán and Blumberg, 2019; Chazal and Michel, 2021) to summarise the CNS in terms of multiscale topological structures (Expert et al., 2019). These topological summaries, the construction of which is described in the methods, can be thought of as multiscale descriptions of the shape of the dataset. Topological summaries, which can be compared by standard methods despite the lack of common reference points, could then be used as the classification features in a support vector machine (SVM). Since the total number of point clouds was relatively small for this type of analysis (Supplementary file 1), we down-sampled each whole CNS point cloud randomly to 8000 points 100 times, producing a total of 3100 point clouds, for each of which we then computed a certain topological summary, called the degree-1 persistence diagram of its alpha complex (Edelsbrunner and Mücke, 1994).

After fixing the necessary hyperparameters, sex classification experiments were run across 5000 random train/test splits of the topological summaries. In each split, the summaries derived from subsamplings of a single CNS point cloud were either all in the training set or all in the testing set, to avoid leaking information. Each time, the SVM was trained once with the animal’s true sex as the target class and once with a randomly assigned sex as the target as a control. We then computed the Pearson’s correlation between the classifier’s output on the testing set and the true (respectively randomised) sex of the animal.

The 5000 splits were used to produce 5000 correlations with the true sex and 5000 correlations with a randomly assigned sex. The distribution of these correlations (Figure 3f), exhibiting clearly that the SVM is able to extract the sex of the animal reliably: only about 1.9% of the splits result in a higher correlation in the control set than in the true data. Moreover, repeating the procedure with simpler point cloud features, like properties of the distributions of inter-nuclei distances, did not produce a significant signal (Figure 3g). Thus, the pattern, which seems hard to describe concisely, is not revealed through simpler descriptors of the neuron configurations, leading us to suspect that CNS sexual dimorphism extends to deeper features of organisation that are both subtle and widely distributed. These results, in addition to the differences in total cell numbers, support sexual dimorphism of the male and female Drosophila CNS at the larval stage.

Potassium channel family member gene expression density in the CNS

Having established a baseline of total numbers of neurons in the larval CNS, we next sought to deploy the quantification pipeline to measure the expression frequency of key neuronal function genes throughout the CNS. We chose to examine the family of voltage-gated potassium channels, which are essential for many aspects of neuronal function and for which Drosophila studies defined the founding members (McCormack, 2003). We generated GAL4 insertions in the Shaker (Sh) (Kv1 family), Shab (Shab) (Kv 2 family), Shaw (Shaw) (Kv3 family), and Shal (Shal) (Kv4 family) (McCormack, 2003) genes using the Trojan/Mimic technique (Diao et al., 2015). As the Sh gene is x-linked, we carried out our quantitation analysis in the male CNS only to avoid potential gene dosage effects. To determine whether our GAL4 reporter lines had patterns of expression consistent with the known properties of these channels, we examined the expression of all four lines in motor neurons, where functional activity for Shaker, Shab, Shaw, and Shal had previously been demonstrated by electrophysiological measurements (Covarrubias et al., 1991; Ryglewski and Duch, 2009). We found that the GAL4 reporters for all 4 channels were expressed as expected in motor neurons (Figure 4—figure supplement 1), consistent with accurate reproduction of the established expression of these proteins.

We next examined the expression frequency of these genes in the entire CNS (Figure 4a–d, Sh-GAL4 Video 11, Shal-GAL4 Video 12, Shab-GAL4 Video 13, Shaw-GAL4 Video 14). We found that Sh and Shal were expressed in large numbers of neurons 8204 ±5.67%, n=10 and 8261 ±3.1%, n=5 respectively, though significantly less (–12.7% and −12.1%, p<0.0001) than the average number of all male neurons (Figure 4a, b and e, Figure 4—source data 1). In contrast, Shab (3057 ±8.21%, n=10) and Shaw (1737 ±4.3%, n=11) were expressed in smaller numbers of neurons (Figure 4c–e), with expression observed in only 32.5% or 18.5% of total male neurons respectively, suggesting more discrete functions within CNS neurons, contrasting with the collective expression of all four genes within motor neurons (Figure 4—figure supplement 1). In particular, Shab and Shaw had very reduced expression in the brain lobes of larva (Figure 4c and d) compared with Sh and Shal (Figure 4a and b). These results establish that our genetic-imaging pipeline can enable quantitation of the expression frequency of families of genes essential for neuronal properties across the entire CNS.

Figure 4 with 1 supplement see all
Quantitation of the number of neurons expressing voltage-gated potassium channel genes.

(a–d) Multiview deconvolved images (left) and z-stack projections (right) (colours represent z position) of potassium channel family members: (a) Sh-GAL4, (b) Shal-GAL4, (c) Shab-GAL4, and (d) Shaw-GAL4. (e) Quantification of the number of labelled nuclei in each line. (a–d) left, scale squares = 50 μm, right, identical magnification, scale bar = 50 μm. (e) Bars indicate minimum and maximum values.

Video 11
Larval CNS labelled with Sh-GAL4.
Video 12
Larval CNS labelled with Shal-GAL4.
Video 13
Larval CNS labelled with Shab-GAL4.
Video 14
Larval CNS labelled with Shaw-GAL4.

Discussion

Establishing the number and identity of cells in the CNS is a foundational metric upon which to construct molecular, developmental, connectomic, and evolutionary atlases of central nervous systems across species (Lent et al., 2012; Devor et al., 2013). Here, we develop and deploy a methodological pipeline to label discrete cell types in the intact Drosophila CNS with genetic reporters designed to facilitate the subsequent segmentation and automated quantification of cell types, in addition to capturing positional coordinates of relative nucleus positions throughout the organ. Using this toolset, we find fewer active neurons, as defined by expression of synaptic protein genes, in the Drosophila larval CNS than most previous predictions and also substantially more glia. We additionally discover previously unsuspected differences in both neuron and glial density and CNS topology between the sexes at the larval stage, when external sex organs are absent, with females possessing both more neurons and fewer glia than males. Topological analysis of point clouds derived from neuronal nucleus position, which can detect potentially subtle and complex geometric structure in the data, also strongly support the existence of differences between the male and female CNS. In addition, deploying these tools, we find that while all members of the Drosophila voltage-gated potassium channel family are expressed in motor neurons, consistent with prior mutant analyses, the Kv2 channel Shab and Kv3 channel Shaw are expressed in a much smaller number of neurons in the CNS than the Kv1 channel Shaker and the Kv4 channel Shal, suggesting conclusions drawn about the coordinated activity of these channels from studies of motor neurons may not be broadly applicable across the CNS, where the genes encoding these channels are frequently not co-expressed.

A number of semiquantitative methods have been employed to estimate the number of neurons in the brains of humans and model organisms, including Drosophila (Lent et al., 2012; Keller et al., 2018). For example, the number of neurons or other cells in the brain has been estimated using stereological counting of subregions. A major limitation of this approach is the assumption of homogenous cell density across the organ or within subregions, which is not supported by the high variability of counts even between samples of similar regions, and thus likely introduces large errors (von Bartheld et al., 2016; Keller et al., 2018). Rough extrapolation of neuronal counts of electron microscope volumes of regions of the Drosophila larval CNS had suggested an estimate of ~15,000 neurons (Meinertzhagen, 2018; Eschbach and Zlatic, 2020). An alternate approach is isotropic fractionation, where all cells in large regions or the entire CNS are dissociated to produce a homogeneous single-cell suspension. Nuclei in the suspension can then be labelled by immunohistochemistry and cells in a subvolume counted in a Neubauer chamber to estimate the total number of cells present. Limitations of the approach include the necessity to ensure complete dissociation of cells while avoiding tissue loss, the requirement for homogenous antibody labelling, and highly accurate dilution (Deniz et al., 2018). This approach has recently been used to estimate the total number of neurons and glia in the adult Drosophila brain and suggested a number of 199,000 neurons (Raji and Potter, 2021), twice prior estimates (Scheffer and Meinertzhagen, 2019; Allen et al., 2020). In contrast to our results in the larval CNS, this study found no significant differences in the number of neurons between the sexes and also found that ‘non-neuronal’ cells, which should include glia, accounted for less than 9% of the total cells counted. In addition to the inherent inaccuracy of the isotropic fractionation technique, which the authors both observed and acknowledge (Raji and Potter, 2021), their use of anti-Elav antibody labelling, which can label some glia in addition to neurons (Berger et al., 2007), or perhaps differences in life stage, may explain some of the discrepancies between our results.

An unpredicted result from our whole CNS neuron quantitation was substantial differences in neuron and glial numbers between the sexes in larva. In adult Drosophila, sexually dimorphic neural circuitry has been observed in the olfactory system (Kimura et al., 2005), and human females have also been reported to have more olfactory bulb neurons and glia than males (Oliveira-Pinto et al., 2014). While sex-specific behavioural differences are obvious in adult Drosophila (Jazin and Cahill, 2010), few sexually dimorphic behavioural differences have been reported in larva (Aleman-Meza et al., 2015). However, male and female larva do differ in nutritional preference (Rodrigues et al., 2015; Davies et al., 2018), which could potentially account for some aspects of the dimorphism we observe. In addition to differences in total cell numbers, our topological methods, which take into account multiscale structure, suggest that differences in CNS structure between the sexes are both subtle (in the mathematical sense) and non-localised in nature, and indeed are not observable with simpler methods of analysis of CNS organisation.

In addition to enabling precise counting of genetically labelled cells, our method allows the relative measurement of discrete cell types or gene expression frequencies throughout the CNS. For example, the relative frequency of glial cells to neurons in the human brain has been long debated (von Bartheld et al., 2016) and in the adult Drosophila brain it has been suggested there are 0.1 glial per neuron (Kremer et al., 2017; Scheffer and Meinertzhagen, 2019; Raji and Potter, 2021). In the larval Drosophila CNS, we found closer to 0.4 glial cells per neuron on average, more similar to the glial-neuron ratios reported for rodents or rabbits (Verkhratsky and Butt, 2018). An important potential caveat, however, is that the large relative ratio of glia we observe in the third instar larva could conceivably be glia produced in advance of adult CNS development. As adult specific neuron numbers expand during pupation, the relative ratio of glia could potentially decline. Additional glial-neuron ratio measurements in the adult CNS will be required to examine this possibility.

Our approach may also allow the assignment of potential functional classes of neurons. For example, from our examination of voltage-gated potassium channel family gene expression, all these channels are collectively expressed in motor neurons; however, the Shab and Shaw genes have more discrete expression patterns in other CNS neuron classes, potentially imbuing these neurons with unique functional characteristics (Chow and Leung, 2020). Future multiplexing of binary genetic expression systems and reporters (Simpson, 2009; del Valle Rodríguez et al., 2011; Diao et al., 2015) should enable neurons or glia to be further quantitively subclassified by gene expression features throughout the entire intact CNS.

Materials and methods

Drosophila stocks

Request a detailed protocol

The following stocks were employed - y[1] w[*]; Mi{y[+mDint2]=MIC}Syt1[MI02197] (BDSC#35973) (Venken et al., 2011), y(1) w(*) Mi(y[+mDint2]=MIC) Sh(MI10885) (BDSC#56260), y(1) w(*);Mi(y[+mDint2]=MIC)Shal(MI10881) (BDSC#56089) (Venken et al., 2011), y(1) w(*); Mi(y[+mDint2]=MIC) Shab(MI00848) (BDSC#34115) (Venken et al., 2011), nSyb-GAL4(GMR57C10)(BDSC#39171) (Pfeiffer et al., 2008), repo-GAL4 (BDSC#7415) (Sepp et al., 2001), repo-QF2 (BDSC#66477) (Lin and Potter, 2016), Shaw-GAL4 (BDSC#60325) (Venken et al., 2011; Li-Kroeger et al., 2018), Ddc-GAL4(BDSC#7009) (Feany and Bender, 2000), TH-GAL4(BDSC#8848) (Friggi-Grelin et al., 2003b), Trh-GAL4(BDSC#38389) (Alekseyenko et al., 2010), UAS_H2A-GFP (Sherer et al., 2020), QUAS_H2B-mCherry (Sherer et al., 2020), brp-GAL4 (this manuscript), UAS_H2A::GFP-T2A-mKok::Caax (this manuscript). All lines were raised on standard media at 25°C, 50% RH.

Generation of brp-GAL4 exon 2 insertion line

Request a detailed protocol

A GAL4.2 sequence was inserted in genome, immediately after the start codon of the Brp-RD isoform using CRISPR based gene editing employing the following constructs. brp gRNA pCDF3: two gRNA sequences targeting each side of the insertion location in exon 2 of brp, were selected using the FlyCRISPR algorithm (http://flycrispr.molbio.wisc.edu/), consisting of 20 nucleotides each (PAM excluded), and predicted to have minimal off-targets. Each individual 20-nucleotide gRNA sequence were inserted into pCFD3 plasmid (Addgene #49,410) using the KLD enzyme mix (New England Biolabs). brp-GAL4 insertion construct: the seven following PCR amplified fragments were assembled using HIFI technology: (1) 1198 bp homology arm covering 5’ UTR until 5’ target site; (2) the region between 5’ target site and the start codon were amplified from Drosophila nos-cas9 (attp2) genomic DNA (a modified Pam sequence was inserted using overlapping primers); (3) Linker-T2A-GAL4.2 sequence was amplified from pBID-DSCP-G-GAL4 (Wang et al., 2012) (the linker-T2A sequence was added upstream of the forward primer); (4) P10-3’UTR was amplified from pJFRC81-10XUAS-IVS-Syn21-GFP-p10 (Addgene 36432); (5) 3xP3-Hsp70pro-dsRed2-SV40polyA selection cassette, flanked by two LoxP sites, was amplified from pHD-sfGFP scareless dsRed (Addgene 80811); (6) The region covering the end of DsRed cassette until 3’ target site; and (7) the 1079 bp homology arm two covering from the 3’ target site to exon 2, were amplified from Drosophila nos-cas9 (attp2) genomic DNA. Full length assembly was topo cloned in zero-blunt end pCR4 vector (Invitrogen), all constructs have been verified by sequencing (Microsynth AG, Switzerland) and injections were carried out into a nos-cas9 (attp2) strain (Ren et al., 2013). Correct insertion of GAL4 was verified by genome sequencing. All primer sequences are included in Appendix 1—key resources table.

Construction of UAS_H2A::GFP-T2A-mKok::Caax

Request a detailed protocol

PCR amplifications were performed using Platinium Superfi polymerase (Invitrogen). The three PCR fragments were assembled together using Hifi technology (Invitrogen): (1) Histone2A (H2A) cDNA was amplified from pDESTP10 LexO-H2A-GFP template (Gift from Steve Stowers) with a synthetic 5’UTR sequence (syn21) added upstream to H2A on the forward primer; (2) sfGFP was amplified from template pHD-sfGFP Scareless dsRed (Addgene 80811); and (3) mKok amplified from pCS2 +ChMermaid S188 (Addgene 53617) with the CAAX membrane tag sequence (Sutcliffe et al., 2017) added at the 3’ end of the protein using the reverse primer. A Thosea asigna virus 2 A(T2A) self-cleaving peptide sequence (Diao et al., 2015), was inserted between sfGFP and mKok, using sfGFP reverse and mKok forward overlapping primers. The full length assembly was TOPO cloned into pCR8GW-TOPO vector (Invitrogen) generating pCR8GW-H2A::GFP-T2A-mKok::Caax. The insert, H2A::GFP-T2A-mKok::Caax was, then, transferred to pBID_UASC_G destination vector (Wang et al., 2012) using LR II clonase kit (Invitrogen) to generate pBID_UAS-H2A::GFP-T2A-mKok::Caax. The transgene was generated by injection into the JK66B landing site. All primer sequences are included in Appendix 1—key resources table.

Generation of novel Trojan GAL4 lines

Request a detailed protocol

MiMIC lines generated by the group of Hugo Bellen (Venken et al., 2011) were acquired from the Bloomington Stock Center. Conversion of Mimic lines to Trojan GAL4 lines was performed as described previously (Diao et al., 2015).

Larval CNS preparation and image acquisition

Request a detailed protocol

Wandering third instar larvae were dissected in 1 × PBS (Mediatech) and fixed with 4% formaldehyde (Sigma-Aldrich) for 20 mins. 1 × PBS were added to remove the fixative, and then the CNS was dissected (Hafer and Schedl, 2006) and rinsed with 1 × PBS with 4% Triton-X 100 for 2 days at 4°C. After rinses, the CNS was embedded in 1% low melting temperature agarose (Peq gold) mixed with 200 nm red fluorescent beads (1:50,000), then introduced into a glass capillary and positioned well separated from each other. After solidification of the agarose, the capillary was mounted to sample holder, transferred to a Zeiss Lightsheet Z.1 microscope and the samples were extruded from the capillary for imaging. CNS images were acquired with a 20 ×/1.0 Apochromat immersion detection objective and two 10 ×/0.2 illumination objectives at five different views, with 1 µm z-intervals. Voxel resolution was 0.317 um.

Image processing and data analysis

Request a detailed protocol

Collected multiview datasets were registered and fused with the Fiji Multiview Reconstruction plugin (Preibisch et al., 2010; Schindelin et al., 2012). Image datasets after multiview deconvolution were analysed with Vision4D 3.0.0 (Arivis AG). A curvature flow filter was first used to denoise the image dataset. Subsequently, a Blob Finder algorithm (Najman and Couprie, 2003) was applied to detect and segment bright rounded three-dimensional sphere-like structures in the images with 4.5 µm set as the diameter. Segmented objects with volume less than 15 µm3 were removed from analysis by segmentation filter to avoid unspecific signals. Subsequently, the number of nuclei and the x, y, z coordinates of the geometric center of each nucleus were output from Vision4D. Where manual counting was employed (Figure 1 and a randomly selected brp-GAL4 labelled CNS), Vision4D was used to visualise and iteratively proceed through and manually annotate the dataset. Example whole CNS datasets where functional neurons or glia are labelled are available (Jiao and McCabe, 2021a; Jiao and McCabe, 2021b). Raw coordinates of the centre of geometry for the nuclei for whole male and female CNS are available in Supplementary file 1. In two-dimensional representations, Z position is indicated by colour coding using the scheme below.

Z-position colour code employed in 2D representations.

Mathematical analysis

Request a detailed protocol

The topological summaries methods employed have previously been introduced (Edelsbrunner and Harer, 2010; Ghrist, 2014; Rabadán and Blumberg, 2019; Chazal and Michel, 2021). For a motivating example of the principles underlying topological summaries, one could think of pearls forming a necklace. Topological summaries express the global structure of the necklace formed by the relationships between the positions of the individual pearls, but are invariant under translation and rotation of the necklace overall. Two such necklaces have topological summaries that are comparable even if the pearls in one have no relationship to the pearls in the other. It is the global structures—such as its circular shape on a large scale, or bulges on a smaller scale—formed by the relationships of the individual pearls of each one that matter.

We trained a machine learning classifier, specifically an SVM, on the CNS nuclei positions in order to evaluate its power in determining characteristics of the animal from which it was derived. The data encompassed all point clouds generated from all CNS lines (brp-GAL4, Syt1-GAL4, nSyb-GAL4, repo-GAL4, Supplementary file 1). Correlation significance (classification power) is determined by comparing the performance of the SVM on the actual classification task to one where each larva is randomly assigned a class.

Mathematically speaking, the nuclei positions from a single CNS form a point cloud, a finite set of points in R3. A possible, naive approach to SVM feature selection for point clouds would be to consider the mean, variance, or other modes of the distribution of pairwise distances within the cloud. These real-valued features could then be passed through, for example, radial basis function kernels for use in SVMs. We focused on very different kind of features, namely ones obtained from the topology of the point clouds. When the point cloud is of low dimension, such as the three-dimensional point clouds arising from nuclei position data, the following approach is relevant. Let X be a finite point cloud in R3. For any r≥0, we let Xr denote the same point cloud, but with each point replaced by a ball of radius r. As r increases, the sequence formed by the Xr expresses different topological features of X. By topological features, we here mean the presence or absence of multiple connected components, unfilled loops, and unfilled cavities.

The figure below illustrates this process in the case of a synthetic two-dimensional point cloud, but the idea extends to any dimension including whole CNS point clouds. When r is small, Xr is topologically very similar to X=X0, and is essentially a collection of disjoint points. When r is very large, Xr is topologically very similar to X∞, i.e., one giant, featureless blob. As the sequence Xr progresses through the continuum of scales between these two trivial extremes, it undergoes non-trivial topological changes: components merge, and loops form and later get filled. In higher dimensions, cavities of various dimensions likewise form and get filled in.

Illustrative 2-dimensional synthetic point cloud.

A small two-dimensional point cloud X viewed at four different scales 0<a<b<c, forming the filtration X=X0⊂ Xa⊂ Xb⊂ Xc.

In the parlance of TDA, we refer to this appearance and disappearance of topological structures as the birth and death of homology classes in various degrees. We capture the whole life cycle with a mathematical object called the persistent homology of the point cloud, which can be fully described by its persistence diagram, a planar collection of points (labelled by multiplicity), whose coordinates encode the birth and death of homological features. For the filtration in the figure above, the persistence diagram that tracks one-dimensional features (i.e. unfilled loops) contains only a single point with coordinates (x, y). Here, the first coordinate, x, is the radius at which the loop is first formed, and the second coordinate, y, is the radius at which the loop has just been filled in. In the example it is clear that a<x<b<y< c.

As multisets of points in the plane, persistence diagrams are not immediately usable as features for SVMs. One way to vectorise persistence diagrams and thus render them digestible by SVMs is to define kernels based on the diagrams, with the heat kernel (Reininghaus et al., 2015) being an oft-used candidate with nice properties. For persistence diagrams P and Q, the heat kernel can informally be defined by the inner product of two solutions of the heat equation—one with an initial condition defined by P, and the other one defined by Q.

In this analysis, we calculated the persistent homology of the alpha complex (Edelsbrunner and Mücke, 1994) of the point clouds, using GUDHI (The GUDHI Editorial Board, 2019). The heat kernels were computed using RFPKOG (Spreemann, 2021). Only the persistence diagrams in degree 1 were used. Since the number of whole CNS point clouds was relatively small, we subsampled each one randomly to 8000 points 100 times, producing a total of 3100 point clouds. This was done both in order to test the stability of the method and to ensure that the variability in the number of points in each cloud is not the source of any signal.

The hyperparameters involved, i.e., the SVM regulariser and the heat kernel bandwidth, were determined by a parameter search in the following way. Six point clouds from males and six from females were randomly selected. All 100 subsampled versions of each of these 12 constituted a training set, for a total of 1200 training point clouds. The remaining 1900 subsampled point clouds constituted the testing set. The Pearson’s correlation between the gender predicted by the SVM on the testing set and the ground truth was computed for each choice of hyperparameters, and a choice in a stable region with high correlation was selected: a regularisation parameter c=10 in the notation of Pedregosa et al., 2011 and a bandwidth of σ=1/100 in the notation of Reininghaus et al., 2015. For the simple distance distribution features, a similar parameter selection process yielded c=10 and a radial kernel bandwidth of 10^5.

Immunofluorescence and confocal microscopy

Request a detailed protocol

Embryos were collected and staged at 25°C on apple agar plates supplemented with yeast paste. Standard methods were used for dechorionation, removal of the vitelline membrane and fixation (Bashaw, 2010). Embryos were stored in 100% ethanol at –20°C before IHC labelling. Embryos were stained with mouse anti-myc9EH10 (1:100, DSHB), visualised with goat anti mouse Alexa Fluor 488 (1:400, Jackson ImmunoResearch) together with conjugated goat anti HRP Alexa Fluor 647 (1:200, Jackson ImmunoResearch) and mounted in VectaShield (Vector Laboratories). Duel colour Z-stack images of stage 15/16 and late stage 17 embryos were obtained on a CSU-W1 Confocal Scanner Unit (Yokogowa, Japan) using two prime BSI express cameras (Teledyne Photometrics). For motor neuron and adult brain preparations, larval fillets from the third instar larvae, or brains from adults were dissected and fixed with 4% formaldehyde (Sigma-Aldrich) for 20 mins. After fixation, samples were rinsed with 1 × PBS and were washed in PBT overnight at 4°C, and then mounted in Vectashield antifade mounting medium. Z-stack images were obtained with a Leica SP8 upright confocal microscope.

For Deadpan staining, the CNS from the third instar larvae were dissected out and fixed with 4% PFA for 20 mins. After fixation, samples were rinsed with 1 × PBS, and permealised with PBT (1 × PBS + 4% Triton-X 100). Antibody stainings were done in PBT +5% normal goat serum. The dilution for chicken anti-GFP (Abcam) was 1:500, for rat anti-Deadpan (Abcam) was 1:50. Goat anti-chicken Alexa Fluor 488, and goat anti-rat Alexa Fluor 594 secondary antibodies were obtained from ThermoFisher and used at the 1:500 dilutions. The CNS were mounted in VectaShield antifade mounting medium, and imaged using a Leica SP8 upright confocal microscope.

Statistical analysis

Request a detailed protocol

Column statistics analyses were performed using GraphPad Prism 9 (GraphPad Software). For Figure 1, statistical significance was determined by unpaired t-test. For Figures 24, statistical significances were determined by ordinary one-way ANOVA, followed by a Tukey’s honestly significant difference test when multiple comparisons were required. The distribution analysis in Figure 2 were performed using matlab (MathWorks). Distances between nuclei coordinates were calculated in matlab using code available at https://doi.org/10.5281/zenodo.6574838 and plotted as a histogram of distance distribution.

Appendix 1

Key resources table

Appendix 1—key resources table
Reagent type (species) or resourceDesignationSource or referenceIdentifiersAdditional information
strain, strain background (Escherichia Coli)One shot top10InvitrogenCat#: C404010
genetic reagent (D. melanogaster)y[1] w[*]; Mi{y[+mDint2]=
MIC}Syt1[MI02197]
Bloomington Drosophila Stock CenterPMID: 21985007BDSC:35973FLYB: FBal0314405
RRID:BDSC_35973
FlyBase symbol: Mi{y[+mDint2]=
MIC}Syt1[MI02197]
genetic reagent (D. melanogaster)y[1] w[*] Mi{y[+mDint2]=
MIC}Sh[MI10885]
Bloomington Drosophila Stock CenterPMID: 21985007BDSC:56260FLYB:
FBal0297530
RRID:BDSC_56260
FlyBase symbol: Mi{MIC}ShMI10885
genetic reagent (D. melanogaster)y[1] w[*];Mi{y[+mDint2]=
MIC}Shal[MI10881]
Bloomington Drosophila Stock CenterPMID: 21985007BDSC:56089FLYB: FBal0295200
RRID:BDSC_56089
FlyBase symbol: Mi{MIC}ShalMI10881
genetic reagent (D. melanogaster)y[1] w[*]; Mi{y[+mDint2]=MIC} Shab[MI00848]Bloomington Drosophila Stock CenterPMID: 21985007BDSC:34115FLYB: FBal0249123
RRID:BDSC_34115
FlyBase symbol:
Mi{MIC}ShabMI00848
genetic reagent (D. melanogaster)nSyb-GAL4Bloomington Drosophila
Stock CenterPMID: 18621688
BDSC: 39171FBgn0013342
RRID:BDSC_39171
Flybase symbol:
P{GMR57C10-GAL4}
genetic reagent (D. melanogaster)repo-GAL4Bloomington Drosophila
Stock Center PMID: 7926782
BDSC:7415FLYB: FBal0127275
RRID:BDSC_7415
FlyBase symbol: P{GAL4}repo
genetic reagent (D. melanogaster)repo-QF2Bloomington Drosophila Stock Centerdoi:10.1534/
genetics.116.191783
BDSC:66477FLYB: FBal0322908
RRID:BDSC_66477
FlyBase symbol: P{ET-QF2.GU}repo
genetic reagent (D. melanogaster)Shaw-GAL4Bloomington Drosophila
Stock CenterPMID: 21985007
BDSC:60325FLYB: FBal0304243
RRID:BDSC_60325
FlyBase symbol:
GAL4Shaw-
MI01735-TG4.1
genetic reagent (D. melanogaster)Ddc-GAL4Bloomington Drosophila Stock Center
doi:10.1006/
dbio.1994.1261
BDSC:7009FLYB: FBtp0012451
RRID:BDSC_7009
FlyBase symbol:
P{Ddc-GAL4.L}
genetic reagent (D. melanogaster)TH-GAL4Bloomington Drosophila Stock CenterPMID: 12555273BDSC:8848FLYB: FBtp0114847
RRID:BDSC_8848
FlyBase symbol:
P{ple-GAL4.F}3
genetic reagent (D. melanogaster)Trh-GAL4Bloomington Drosophila Stock Centerdoi:10.1371/journal.
pone.0010806
BDSC:38389FLYB:
FBtp0055412
RRID:BDSC_38389
FlyBase symbol:
P{Trh-GAL4.long}
genetic reagent (D. melanogaster)UAS_H2A-GFPSteve Stowers (Montana SU)doi:10.1371/journal.
pgen.1008609
FLYB: FBgn0001196
genetic reagent (D. melanogaster)QUAS_H2A-mCherrySteve Stowers(Montana SU)doi:10.1371/journal.
pgen.1008609
FLYB: FBgn0001196
genetic reagent (D. melanogaster)brp-GAL4This paper, available
upon request, https://www.epfl.ch/labs/mccabelab/resources/
FLYB:FBgn0259246See Materials and
Methods, Section 2
genetic reagent (D. melanogaster)Syt1-Gal4This paper, available
upon request, https://www.epfl.ch/labs/mccabelab/resources/
FLYB: FBal0314405See Materials and
Methods, Section 4
genetic reagent (D. melanogaster)Sh-Gal4This paper, available
upon request,https://www.epfl.ch/labs/mccabelab/resources/
FLYB: FBal0297530See Materials and
Methods, Section 4
genetic reagent (D. melanogaster)Shal-Gal4This paper, available upon request,https://www.epfl.ch/labs/mccabelab/resources/FLYB: FBal0249123See Materials and
Methods, Section 4
genetic reagent (D. melanogaster)Shab-Gal4This paper, available
upon request,https://www.epfl.ch/labs/mccabelab/resources/
FLYB: FBal0249123See Materials and
Methods, Section 4
genetic reagent (D. melanogaster)UAS_H2A::GFP-
T2A-mKok::Caax
This paper, available
upon request,https://www.epfl.ch/labs/mccabelab/resources/
FLYB: FBgn0001196See Materials and
Methods, Section 3
antibodyanti-myc9EH10(Mouse monoclonal)DSHBDSHB Cat# 9E 10, RRID:AB_2266850IF(1:100)
antibodyanti-GFP(Chicken polyclonal)AbcamCat#ab13970IF(1:500)
antibodyanti-Deadpan(Rat monoclonal)AbcamCat#ab195173IF(1:50)
antibodyGoat Anti-Mouse IgG (H+L), Alexa Fluor 488 (Goat polyclonal Secondary Antibody)Jackson
ImmunoResearch
Cat#115-545-166IF(1:400)
antibodyGoat Anti-Horseradish Peroxidase,Alexa Fluor 647 (Goat polyclonal)Jackson
ImmunoResearch
Cat#123-605-021IF(1:200)
antibodyGoat anti-Chicken IgY (H+L), Alexa Fluor 488 (Goat polyclonal Secondary Antibody)ThermoFisherCat#A-11039IF(1:500)
antibodyGoat anti-Rat IgG (H+L) Cross-Adsorbed, Alexa Fluor 594 (Goat polyclonal Secondary Antibody)ThermoFisherCat#A-11007IF(1:500)
recombinant DNA reagentpBID_DSCP-G-Gal4 (plasmid)McCabe Lab,Available
upon request.https://www.epfl.ch/labs/mccabelab/resources/
Cat# #35,200
recombinant DNA reagentpJFRC81-10XUAS-IVS-
Syn21-GFP-p10 (plasmid)
AddgeneCat#: 36,432
recombinant DNA reagentpHD-sfGFP Scareless dsRed (plasmid)AddgeneCat#: 80,811
recombinant DNA reagentpCFD3 (plasmid)AddgeneCat#: 49,410
recombinant DNA reagentpCR4 brp-Gal4This paper, available
upon request, https://www.epfl.ch/labs/mccabelab/resources/
CRISPR construct inserted
in D. Melanogaster
recombinant DNA reagentpBID LexO_H2A-mCherryGift from Steve Stowers
(Montana SU),
DOI: 10.1371/journal.pgen.1008609
recombinant DNA reagentpCS2+ChMermaid S188AddgeneCat#: 53,617
recombinant DNA reagentpBID-UAS_H2A::GFP-T2A-mKok::CaaxThis paper, available
upon request, https://www.epfl.ch/labs/mccabelab/resources/
construct inserted
in D. Melanogaster
sequence-based reagentpCFD3 gRNA brp-5’This paperGuide RNA for Brp CRISPR knock inGGTGAACCGA
CCGGGACAAC
sequence-based reagentpCFD3 gRNA brp-3’This paperGuide RNA for Brp CRISPR knock inGGGAGCCCCGC
GACCGCTCC
sequence-based reagentbrp Ha1 FoThis paperPCR primerGAGAGAGCATCT
CGATTGTGCCGTGTG
sequence-based reagentbrp Pam7 ReThis paperPCR primerAATGTTGTCCCG
GTCGGTTCACCG
sequence-based reagentbrp Pam7_In1 ReThis paperPCR primerTTCTAGCGTCCAA
CGGCTCAGCTGTG
GGCCATTTTCTAGT
AATGTTGTCCCGG
TCGGTTCACCG
sequence-based reagentbrp HA_In1 FoThis paperPCR primerACTAGAAAATGGCC
CACAGCTGAGCC
sequence-based reagentbrp V5_In1 ReThis paperPCR primerTAGAATCGAGACCG
AGGAGAGGGTTAGGG
ATAGGCTTACCCATT
GCTGAAATTCACACA
CACACAGAATTCATGAG
sequence-based reagentbrp V5_ FoThis paperPCR primerGGTAAGCCTATCCC
TAACCCTCTCCTC
sequence-based reagentbrp PB5’ ReThis paperPCR primerTTAAGGGATCTTTCTA
TTAGTATAACACTGCATGC
sequence-based reagentbrp Ex2 foThis paperPCR primerAAATTGCATGCAGTGTT
ATACTAATAGAAAGATCC
CTTAATCGGCAGTCCAT
ACTACCGCGACATGGATG
sequence-based reagentbrp Pam2_ReThis paperPCR primerTCTGGAGCGGT
CGCGGGGC
sequence-based reagentbrp Pam2_Brp Ha2 ReThis paperPCR primerGCTCGTCCTCTAGGTAC
AGGCCCCGTTCGAGGGA
TCTGTCTCTGGAGC
GGTCGCGGGG
sequence-based reagentbrp Ha2 FoThis paperPCR primerGACAGATCCCTC
GAACGGGGCC
sequence-based reagentSyn21 H2A FoThis paperPCR primerAACTTAAAAAAAAAA
ATCAAAATGTCTGGA
CGTGGAAAAGGTGGC
sequence-based reagentH2A ReThis paperPCR primerCCCAAGAAGACC
GAGAAGAAGGCC
sequence-based reagentH2A-GFP FoThis paperPCR primerACAGGCTGTTCTGT
TGCCCAAGAAGACC
GAGAAGAAGGCCAT
GGTGTCCAAGGG
CGAGGAG
sequence-based reagentGFP-T2A ReThis paperPCR primerGGGTTCTCCTCCAC
ATCGCCGCAGGTCAG
CAGGCTGCCGCGGC
CCTCCTTGTACAGCT
CATCCATGCCCAGG
sequence-based reagentT2A-mKok FoThis paperPCR primerGCGGCAGCCTGCT
GACCTGCGGCGATG
TGGAGGAGAACCCC
GGGCCCATGGTGAGT
GTGATTAAACCAG
AGATGAAGATG
sequence-based reagentmKok-Caax ReThis paperPCR primerTTACATAATTACACA
CTTTGTCTTTGACTT
CTTTTTCTTCTTTTTA
CCATCTTTGCTCATGG
AATGAGCTACTGCAT
CTTCTACCTGC
chemical compound, drugFormaldehyde 37%SigmaCat#: 252,549
chemical compound, drugLow melt agarosePeq goldCat#: 35–2010
chemical compound, drugVECTASHIELD
Antifade Mounting Media
VECTOR LaboratoriesCat#: H-1000
chemical compound, drugFluoSpheres, 0.2 µm, red fluorescent (580/605)Thermo Fisher ScientificCat#: F8810
commercial assay or kitZero Blunt TOPO
PCR Cloning Kit
InvitrogenCat#: 450,245
commercial assay or kitpCR8/GW/TOPO
TA Cloning Kit
InvitrogenCat#: K250020
commercial assay or kitGateway LR Clonase
II Enzyme mix
InvitrogenCat#:11791020
commercial assay or kitNEBuilder HiFi
DNA Assembly Master Mix
New England BiolabsCat#: E2621S
commercial assay or kitKLD enzyme mixNew England BiolabsCat#:M0554S
commercial assay or kitPlatinium
Superfi polymerase
InvitrogenCat#:12359010
software, algorithmCNS nuclei distance
(code for Matlab)
https://doi.org/10.5281/zenodoThis manuscript
software, algorithmFijihttps://fiji.sc/RRID:SCR_002285
software, algorithmArivis Vision4D 3.0.0ArivisRRID:SCR_018000
software, algorithmMATLAB (R2018a)MathWorksRRID:SCR_001622
software, algorithmGraphPad Prism 9.0GraphPadRRID:SCR_002798

Data availability

All data generated or analysed during this study are included in the manuscript and supporting files. A source data file has been provided for all figures. Raw co-ordinates of the center of geometry of CNS nuclei for all animals is attached as Supplementary file 1. Matlab code employed is available at https://doi.org/10.5281/zenodo.6574838. Example unprocessed whole CNS microscopy data is publicly available for neurons - https://doi.org/10.5281/zenodo.5585334 and for glia https://doi.org/10.5281/zenodo.5585358.

The following data sets were generated
    1. McCabe B
    2. Jiao W
    (2021) Zenodo
    Whole Brain Drosophila Larval Neurons.
    https://doi.org/10.5281/zenodo.5585334
    1. McCabe B
    2. Jiao W
    (2021) Zenodo
    Whole Brain Drosophila Larval Glia.
    https://doi.org/10.5281/zenodo.5585358

References

    1. Deitcher DL
    2. Ueda A
    3. Stewart BA
    4. Burgess RW
    5. Kidokoro Y
    6. Schwarz TL
    (1998)
    Distinct requirements for evoked and spontaneous release of neurotransmitter are revealed by mutations in the Drosophila gene neuronal-synaptobrevin
    The Journal of Neuroscience 18:2028–2039.
  1. Book
    1. Ghrist R
    (2014)
    Elementary Applied Topology
    Seattle: Createspace.
    1. Li H
    2. Janssens J
    3. De Waegeneer M
    4. Kolluru SS
    5. Davie K
    6. Gardeux V
    7. Saelens W
    8. David FPA
    9. Brbić M
    10. Spanier K
    11. Leskovec J
    12. McLaughlin CN
    13. Xie Q
    14. Jones RC
    15. Brueckner K
    16. Shim J
    17. Tattikota SG
    18. Schnorrer F
    19. Rust K
    20. Nystul TG
    21. Carvalho-Santos Z
    22. Ribeiro C
    23. Pal S
    24. Mahadevaraju S
    25. Przytycka TM
    26. Allen AM
    27. Goodwin SF
    28. Berry CW
    29. Fuller MT
    30. White-Cooper H
    31. Matunis EL
    32. DiNardo S
    33. Galenza A
    34. O’Brien LE
    35. Dow JAT
    36. Jasper H
    37. Oliver B
    38. Perrimon N
    39. Deplancke B
    40. Quake SR
    41. Luo L
    42. Aerts S
    43. Agarwal D
    44. Ahmed-Braimah Y
    45. Arbeitman M
    46. Ariss MM
    47. Augsburger J
    48. Ayush K
    49. Baker CC
    50. Banisch T
    51. Birker K
    52. Bodmer R
    53. Bolival B
    54. Brantley SE
    55. Brill JA
    56. Brown NC
    57. Buehner NA
    58. Cai XT
    59. Cardoso-Figueiredo R
    60. Casares F
    61. Chang A
    62. Clandinin TR
    63. Crasta S
    64. Desplan C
    65. Detweiler AM
    66. Dhakan DB
    67. Donà E
    68. Engert S
    69. Floc’hlay S
    70. George N
    71. González-Segarra AJ
    72. Groves AK
    73. Gumbin S
    74. Guo Y
    75. Harris DE
    76. Heifetz Y
    77. Holtz SL
    78. Horns F
    79. Hudry B
    80. Hung R-J
    81. Jan YN
    82. Jaszczak JS
    83. Jefferis GSXE
    84. Karkanias J
    85. Karr TL
    86. Katheder NS
    87. Kezos J
    88. Kim AA
    89. Kim SK
    90. Kockel L
    91. Konstantinides N
    92. Kornberg TB
    93. Krause HM
    94. Labott AT
    95. Laturney M
    96. Lehmann R
    97. Leinwand S
    98. Li J
    99. Li JSS
    100. Li K
    101. Li K
    102. Li L
    103. Li T
    104. Litovchenko M
    105. Liu H-H
    106. Liu Y
    107. Lu T-C
    108. Manning J
    109. Mase A
    110. Matera-Vatnick M
    111. Matias NR
    112. McDonough-Goldstein CE
    113. McGeever A
    114. McLachlan AD
    115. Moreno-Roman P
    116. Neff N
    117. Neville M
    118. Ngo S
    119. Nielsen T
    120. O’Brien CE
    121. Osumi-Sutherland D
    122. Özel MN
    123. Papatheodorou I
    124. Petkovic M
    125. Pilgrim C
    126. Pisco AO
    127. Reisenman C
    128. Sanders EN
    129. Dos Santos G
    130. Scott K
    131. Sherlekar A
    132. Shiu P
    133. Sims D
    134. Sit RV
    135. Slaidina M
    136. Smith HE
    137. Sterne G
    138. Su Y-H
    139. Sutton D
    140. Tamayo M
    141. Tan M
    142. Tastekin I
    143. Treiber C
    144. Vacek D
    145. Vogler G
    146. Waddell S
    147. Wang W
    148. Wilson RI
    149. Wolfner MF
    150. Wong Y-CE
    151. Xie A
    152. Xu J
    153. Yamamoto S
    154. Yan J
    155. Yao Z
    156. Yoda K
    157. Zhu R
    158. Zinzen RP
    159. FCA Consortium§
    (2022) Fly Cell Atlas: A single-nucleus transcriptomic atlas of the adult fruit fly
    Science (New York, N.Y.) 375:eabk2432.
    https://doi.org/10.1126/science.abk2432
  2. Book
    1. Najman L
    2. Couprie M
    (2003) Watershed Algorithms and Contrast Preservation In
    In: Nyström I, Baja G, Svensson S, editors. Discrete Geometry for Computer Imagery. Berlin, Heidelberg: Springer. pp. 62–71.
    https://doi.org/10.1007/978-3-540-39966-7_5
    1. Pedregosa F
    2. Varoquaux G
    3. Gramfort A
    4. Michel V
    5. Thirion B
    6. Grisel O
    7. Blondel M
    8. Prettenhofer P
    9. Weiss R
    10. Dubourg V
    11. Vanderplas J
    12. Passos A
    13. Cournapeau D
    (2011)
    Scikit-learn: Machine Learning in Python
    Journal of Machine Learning Research: JMLR 1:2825–2830.
  3. Conference
    1. Reininghaus J
    2. Huber S
    3. Bauer U
    4. Kwitt R
    (2015) A stable multi-scale kernel for topological machine learning
    2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR. pp. 4741–4748.
    https://doi.org/10.1109/CVPR.2015.7299106
  4. Website
    1. The GUDHI Editorial Board
    (2019) GUDHI library
    Accessed June 1, 2021.
    1. White JG
    2. Southgate E
    3. Thomson JN
    4. Brenner S
    (1986) The structure of the nervous system of the nematode Caenorhabditis elegans
    Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences 314:1–340.
    https://doi.org/10.1098/rstb.1986.0056

Decision letter

  1. Sonia Sen
    Reviewing Editor; Tata Institute for Genetics and Society, India
  2. K VijayRaghavan
    Senior Editor; National Centre for Biological Sciences, Tata Institute of Fundamental Research, India
  3. Sonia Sen
    Reviewer; Tata Institute for Genetics and Society, India

Our editorial process produces two outputs: i) public reviews designed to be posted alongside the preprint for the benefit of readers; ii) feedback on the manuscript for the authors, including requests for revisions, shown below. We also include an acceptance summary that explains what the editors found interesting or important about the work.

Decision letter after peer review:

Thank you for submitting your article "Intact Drosophila Whole Brain Cellular Quantitation reveals Sexual Dimorphism" for consideration by eLife. Your article has been reviewed by 2 peer reviewers, including Sonia Sen as Reviewing Editor and Reviewer #1, and the evaluation has been overseen by K VijayRaghavan as the Senior Editor.

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

Essential revisions:

While noting the value and general rigour of this work, we had a few concerns that the authors could address. These are listed below.

1) The new Gal4 lines: These lines form the basis of this study and we anticipate that they will be of value to the broader community. For both these reasons, we request that the authors verify them better, for example, with neuroblast-specific markers and/or co-labels with markers of mature neurons.

2) Comparison with previous studies: We were concerned about the comparison with previous studies – please see detailed reviews below. Could the authors please revisit the text surrounding this, and perhaps provide a systematic comparison of the previous studies that have 'counted' cell numbers?

3) Interpretation of some of the findings: Some of the findings are interesting, but we were not sure what they meant (please see the detailed comments below). Could the authors discuss this? If they feel it would be too speculative, they could perhaps use the Ideas and Speculation section within discussion to do this.

4) Description of methods: Could the authors please describe some of their methods in more detail? Please see the detailed reviews below, but, for example, the procedures leading to the topographical comparisons, or the comparison of 3D coordinates across brains.

Reviewer #1 (Recommendations for the authors):

– The authors have generated invaluable Gal4 lines through their study. We would have liked to have seen a better characterization of them particularly because we think they will be of value to the broader community. For example with neuroblast-specific markers or co-labels with markers of mature neurons.

– It would be nice to have a more systematic description of previous studies that have estimated cell numbers in larval and adult brains. The authors have done this, but the larval and adult studies are listed together, making it hard to assess the different approaches and their advantages and limitations. Could the authors please look into discussing it systematically in this way?

– The authors quote previous studies suggesting glia constitute 10% of the cell in the brain – this is for the adult, isn't it? Significant neurogenesis occurs between early L3 and the adult and so is it possible that the 30% observed in this study transitions to 10%?

– I did not quite understand how one can compare 3D coordinates of nuclei across brains if there isn't a common reference? Could the authors please elaborate on this?

– What do the authors interpret from the nuclear distance distribution patterns of neurons and glia?

– Similarly, the authors say that the sexual dimorphism extends to deeper features, is subtle, but widespread. Could the authors speculate what these subtle, widespread features might be?

Reviewer #2 (Recommendations for the authors):

The conclusions of this paper are supported mainly by data, but some aspects of data analysis and interpretation need to be clarified and extended to solidify their findings.

1) The paper is generally written well, but some data interpretations are based on inaccurate and shaky estimates, which are better avoided. In my understanding, none of the precedents formally publish the number of neurons and glia in the whole nervous system of third instar larvae. As mentioned in the text, there are indeed some estimates that range from 10,000 to 15,000 (P3, L15, Scott, et al., 2001), but they are not published as data and are not of the same age as their animal. Therefore, it is better for the authors to avoid direct comparisons with this wide range of statistics to draw numerical arguments (e.g., P8. L137-138). Instead, those studies collectively suggest the present results are in good agreement with them, solidifying their work rather than contradicting.

2) Usage of the term "brain" has to be strict to avoid any future confusion (cf. title and abstract). While I agree that they found sexual dimorphism in the larval nervous system, I am unsure whether it is about the brain. The central nervous system of larvae consists of the brain and ventral nervous system, and the methodology of the current study analyzes both. Numerically speaking, the latter is the majority (4~5 times more), suggesting that the difference is somewhat related to other peripheral parts of the nervous system than the brain. I suggest the authors revise the relevant points and put their work in the proper context.

3) Text writing in the methodological part could work better to substantiate their main findings further. In particular, the procedures leading to the topographical comparisons that contribute to their main conclusion (P8. 4th paragraph – P9. 2nd paragraph) uses many unfamiliar mathematical terms without proper explanations that make it difficult for non-specialists to follow.

4) In the distribution of inter-nuclei distances (Figure 2e), I see fractions of the cells fall into the range of distance 0, suggestive of double counting. Some questions about nuclei counting accuracy.

5) I would like to know the specific citation behind their notion that "genes likely to be expressed only in active larval neurons with synaptic connections but not by neuroblasts or by immature neurons". Brp is known to be expressed in the developing embryo (Wagh et al., Neuron, 2006).

6) (P9. L192) "As the Sh gene is x-linked, we carried out our quantification analysis in male brains only to avoid potential gene dosage effects."

How does the dosage effect affect their cell counting? In other words, is this data appropriate being here without contributing the main conclusion of sexual dimorphism?

7) A list of questions for methodological clarifications.

– What genotypes were used for the topographical comparisons (Figure 3f 3g)?

– (P8. L160) What are the "multi-scale topological structures"?

– (P8. L162-164) I do not follow this part. The authors need to elaborate on the explanation.

– What is the voxel resolution of their images?

– How are inter-nuclei distances calculated (Figure 2e)?

– (P16. L356) How is the volume cutoff (15um3) determined?

– What is coded by the color in Figure 2a-2d and others?

– What is the unit of the abscissa in Figure 2e?

– (P8. L137) What is the denominator to get the value (~15 to 30%)?

8) (Figure 1, 2, 3, 4) GeneName>GAL4. Is this the standard way to describe the genotype?

https://doi.org/10.7554/eLife.74968.sa1

Author response

Reviewer #1 (Recommendations for the authors):

– The authors have generated invaluable Gal4 lines through their study. We would have liked to have seen a better characterization of them particularly because we think they will be of value to the broader community. For example with neuroblast-specific markers or co-labels with markers of mature neurons.

We thank the reviewer for their kind and insightful comments. To address their concern, we have carried out several additional experiments to further characterise these lines. We first examined the expression of these lines during embryo development (new Figure 2—figure supplement 2). We observed no expression at stage 15-16 or stages prior to this (Figure 2—figure supplement 2a). However, at stage 17 all three lines had neuronal expression (Figure 2—figure supplement 2a). This is in good agreement with the onset of synaptic activity in embyros (Baines and Bate 1998). We also examined the expression of Deadpan, a transcription factor that labels neuroblast nuclei, in the larval CNS together with histoneGFP expression labelling driven by all three neuronal lines. Consistent with our lines not labelling neuroblasts, we found that the expression of brp-Gal4, Syt1-Gal4 and nSyb-Gal4 were mutually exclusive from Deadpan labelling (Figure 2—figure supplement 2b). Finally, we also examined the expression of all three lines in the adult CNS and found that all three also appear to express in neurons in the adult brain (Figure 2—figure supplement 2c).

– It would be nice to have a more systematic description of previous studies that have estimated cell numbers in larval and adult brains. The authors have done this, but the larval and adult studies are listed together, making it hard to assess the different approaches and their advantages and limitations. Could the authors please look into discussing it systematically in this way?

As correctly described by reviewer #2, most of these prior estimates seem to be rough approximations and are not based on data or at least not explicitly linked to data in the source manuscripts. We have however rewritten this section as the reviewer requested. Larval and adult estimates are now separated and additional references added to support the range of numbers proposed (lines 16-23 revised manuscript).

– The authors quote previous studies suggesting glia constitute 10% of the cell in the brain – this is for the adult, isn't it? Significant neurogenesis occurs between early L3 and the adult and so is it possible that the 30% observed in this study transitions to 10%?

This is an excellent point. We have added this potential explanation of divergence in our measurements to this discussion (line 305-309 revised manuscript).

– I did not quite understand how one can compare 3D coordinates of nuclei across brains if there isn't a common reference? Could the authors please elaborate on this?

Reviewer #2 also asked for clarification about this novel biological data analysis method. Topological summaries indeed lack common reference points but nonetheless can be compared by standard methods. They can be thought of as multi-scale descriptions of the shape of the data set. In addition to adding several basic references to topological data analysis to aid biologists, we added this ‘motivating example’ or thought experiment to help illustrate the underlying principle to the mathematics methods section (revised manuscript lines 401-409)

“For a motivating example of the principles underlying topological summaries, one could think of pearls forming a necklace. Topological summaries express the global structure of the necklace formed by the relationships between the positions of the individual pearls, but are invariant under translation and rotation of the necklace overall. Two such necklaces have topological summaries that are comparable even if the pearls in one have no relationship to the pearls in the other. It is the global structures – such as its circular shape on a large scale, or bulges on a smaller scale – formed by the relationships of the individual pearls of each one that matter”

– What do the authors interpret from the nuclear distance distribution patterns of neurons and glia?

We interpret these results as yet another confirmation that cells labelled are different between the neuronal Gal4 lines and the glial lines. We have edited the manuscript to make this point explicit (line 135-137 revised manuscript)

– Similarly, the authors say that the sexual dimorphism extends to deeper features, is subtle, but widespread. Could the authors speculate what these subtle, widespread features might be?

We are very reluctant to speculate as we have no data on which to propose a hypothesis. Topology analysis is telling us the differences between the female and male CNS is subtle, widespread and clearly detectible but also difficult to identify with simpler analysis approaches. An interesting future direction might be to examine CNS gynandromorphs to determine if these differences could be more discreetly localised.

Reviewer #2 (Recommendations for the authors):

The conclusions of this paper are supported mainly by data, but some aspects of data analysis and interpretation need to be clarified and extended to solidify their findings.

We thank the reviewer for their supportive comments and have sought to address their concerns.

1) The paper is generally written well, but some data interpretations are based on inaccurate and shaky estimates, which are better avoided. In my understanding, none of the precedents formally publish the number of neurons and glia in the whole nervous system of third instar larvae. As mentioned in the text, there are indeed some estimates that range from 10,000 to 15,000 (P3, L15, Scott, et al., 2001), but they are not published as data and are not of the same age as their animal. Therefore, it is better for the authors to avoid direct comparisons with this wide range of statistics to draw numerical arguments (e.g., P8. L137-138). Instead, those studies collectively suggest the present results are in good agreement with them, solidifying their work rather than contradicting.

This point was also raised Reviewer #1. The reviewer is correct that previous estimates were not published as data and have a broad range. Nonetheless they have been widely cited with the most recent estimates in higher range ~15K (e.g. Meinertzhagen, 2018; Eschbach and Zlatic, 2020).

We have rewritten this section (lines 13-23 revised manuscript) to make the point that the cited numbers are approximations more explicit and revised our conclusions as the reviewer suggested to indicate our data supports neuron numbers at the lowest end of previous estimates. As the reviewer suggested, we have removed all direct numerical comparisons to previous approximations.

2) Usage of the term "brain" has to be strict to avoid any future confusion (cf. title and abstract). While I agree that they found sexual dimorphism in the larval nervous system, I am unsure whether it is about the brain. The central nervous system of larvae consists of the brain and ventral nervous system, and the methodology of the current study analyzes both. Numerically speaking, the latter is the majority (4~5 times more), suggesting that the difference is somewhat related to other peripheral parts of the nervous system than the brain. I suggest the authors revise the relevant points and put their work in the proper context.

The reviewer is absolutely correct. We have edited the title, abstract, text and figure legends to use the term ‘central nervous system (CNS)’ when appropriate, which is the case for the majority of our experiments. We make explicit that the Drosophila CNS and our experiments include measurements of the brain and VNC combined (line 13-16 revised manuscript).

3) Text writing in the methodological part could work better to substantiate their main findings further. In particular, the procedures leading to the topographical comparisons that contribute to their main conclusion (P8. 4th paragraph – P9. 2nd paragraph) uses many unfamiliar mathematical terms without proper explanations that make it difficult for non-specialists to follow.

Indeed, there are many terms here that may be unfamiliar to those of us outside mathematics fields. To address the reviewers concern, we have made a number of edits. In the main text, we have added additional accessible references to serve as general primers for non-specialists and added an additional explanation of the general approach (lines 177-183 revised manuscript). We have also expanded the methods section to include further general topology references and a ‘motivating example’ or thought experiment to help illustrate the underlying principle given that this type of topological data analysis is new for neuroscience microscopy data (lines 401-409 revised manuscript).

4) In the distribution of inter-nuclei distances (Figure 2e), I see fractions of the cells fall into the range of distance 0, suggestive of double counting. Some questions about nuclei counting accuracy.

The plot in Figure 2e is derived from a regression analysis (the raw inter-nuclei distance values are provided in supplementary data1 and the regression analysis matlab code is publicly available at https://doi.org/10.5281/zenodo.6574838). As the reviewer correctly points out the input data are distances, so we should have should have set the range of the analysis outputs to positive values. We have rerun the analysis making this correction and edited the text and figure.

5) I would like to know the specific citation behind their notion that "genes likely to be expressed only in active larval neurons with synaptic connections but not by neuroblasts or by immature neurons". Brp is known to be expressed in the developing embryo (Wagh et al., Neuron, 2006).

This point was also raised Reviewer #1. To address this concern, we have carried out several additional experiments to further characterise these lines. We first examined the expression of these lines during embryo development (new Figure 2—figure supplement 2). We observed no expression at stage 15-16 or stages prior to this (Figure 2—figure supplement 2a). However, at stage 17 all three lines had expression (Figure 2—figure supplement 2a). This is in good agreement with the onset of synaptic activity in embyros (Baines and Bate 1998).

We have edited the statement cited to read ‘genes likely to be expressed only in active neurons with synaptic connections but not by neuroblasts or by immature neurons’ (line 90 revised manuscript) to define our rationale and subsequently confirm these lines indeed initiate expression when synaptic activity begins (new Figure 2—figure supplement 2) and lines 110-121 in the revised manuscript.

6) (P9. L192) "As the Sh gene is x-linked, we carried out our quantification analysis in male brains only to avoid potential gene dosage effects."

How does the dosage effect affect their cell counting? In other words, is this data appropriate being here without contributing the main conclusion of sexual dimorphism?

Our goal with this series of experiments was to illustrate the utility and application of the technique to investigate the frequency of expression of a gene family throughout the CNS and we believe we have made some surprising observations into the relative abundance of K+ channels. Our concern was that well documented X-linked gene dosage effects could influence expression of the Sh-Gal4 line (which in on the X chromosome) which is why we choose to carry out these experiments in only one sex. We think these experiments are useful to illustrate another utility of the technique independently of other findings related to sexual dimorphism in the manuscript. We have edited the abstract to make this additional goal clear.

7) A list of questions for methodological clarifications.

– What genotypes were used for the topographical comparisons (Figure 3f 3g)?

All point clouds from all CNS lines were used. This has been clarified to the methods (revised manuscript line 413)

– (P8. L160) What are the "multi-scale topological structures"?

We have added additional references that explain these terms, see also answer to question 3 above.

– (P8. L162-164) I do not follow this part. The authors need to elaborate on the explanation.

We have added additional explanation in the main text and methods, see also answer to question 3 above.

– What is the voxel resolution of their images?

Our voxel resolution is 0.317um. We had added this to the methods (revised manuscript line 383)

– How are inter-nuclei distances calculated (Figure 2e)?

Arivis4D outputs the three dimensional coordinates of the geometric center of every labelled nuclei. These co-ordinates were then processed using a Matlab code we generated to calculate internuclei the distance in 3D volume. This code is publicly available to download at https://doi.org/10.5281/zenodo.6574838 and this has been added to the revised methods.

– (P16. L356) How is the volume cutoff (15um3) determined?

The volume was calculated by as the component of the radius of a the nucleus -1.5um producing a spherical volume – 15um3 (V=4-3×πr!). This was empirically validated by volume measurements of subsets of neurons. This has been clarified in the methods.

– What is coded by the color in Figure 2a-2d and others?

Colour codes Z-coordinate position in these 2D images. We have explained this in the revised figure legends and added the colour scale to the methods (line 399).

– What is the unit of the abscissa in Figure 2e?

um. We have corrected this error.

– (P8. L137) What is the denominator to get the value (~15 to 30%)?

We have removed this statement and calculation as described above based on the reviewers valid criticism.

8) (Figure 1, 2, 3, 4) GeneName>GAL4. Is this the standard way to describe the genotype?

Unfortunately, there is no ‘standard’ way to describe these shorthand genotypes (full genotypes are included in the methods and key resources file). A -,. , _ or > have been used to describe a promoter upstream of a Gal4 in various manuscripts. However, as suggested, we have replaced > with – as this seems the most widely employed shorthand.

References

Baines RA, Bate M. 1998. Electrophysiological Development of Central Neurons in the Drosophila Embryo. J Neurosci 18:4673–4683. doi:10.1523/JNEUROSCI.18-12-04673.1998

https://doi.org/10.7554/eLife.74968.sa2

Article and author information

Author details

  1. Wei Jiao

    Brain Mind Institute, EPFL - Swiss Federal Institute of Technology, Lausanne, Switzerland
    Contribution
    Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Resources, Validation, Visualization, Writing – original draft, Writing – review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-1536-5937
  2. Gard Spreemann

    Brain Mind Institute, EPFL - Swiss Federal Institute of Technology, Lausanne, Switzerland
    Present address
    Telenor Research, Fornebu, Norway
    Contribution
    Conceptualization, Investigation, Methodology, Software, Validation, Visualization, Writing – original draft, Writing – review and editing
    Competing interests
    No competing interests declared
  3. Evelyne Ruchti

    Brain Mind Institute, EPFL - Swiss Federal Institute of Technology, Lausanne, Switzerland
    Contribution
    Resources, Visualization
    Competing interests
    No competing interests declared
  4. Soumya Banerjee

    Brain Mind Institute, EPFL - Swiss Federal Institute of Technology, Lausanne, Switzerland
    Contribution
    Resources
    Competing interests
    No competing interests declared
  5. Samuel Vernon

    Brain Mind Institute, EPFL - Swiss Federal Institute of Technology, Lausanne, Switzerland
    Contribution
    Investigation
    Competing interests
    No competing interests declared
  6. Ying Shi

    Brain Mind Institute, EPFL - Swiss Federal Institute of Technology, Lausanne, Switzerland
    Contribution
    Formal analysis, Methodology
    Competing interests
    No competing interests declared
  7. R Steven Stowers

    Department of Microbiology and Cell Biology, Montana State University, Bozeman, United States
    Contribution
    Resources, Writing – review and editing
    Competing interests
    No competing interests declared
  8. Kathryn Hess

    Brain Mind Institute, EPFL - Swiss Federal Institute of Technology, Lausanne, Switzerland
    Contribution
    Methodology, Supervision, Writing – original draft, Writing – review and editing
    Competing interests
    No competing interests declared
  9. Brian D McCabe

    Brain Mind Institute, EPFL - Swiss Federal Institute of Technology, Lausanne, Switzerland
    Contribution
    Conceptualization, Funding acquisition, Project administration, Supervision, Writing – original draft, Writing – review and editing
    For correspondence
    brian.mccabe@epfl.ch
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-1620-0501

Funding

Swiss National Science Foundation (31003A_179587)

  • Brian D McCabe

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

This work was supported in part using the resources and services of the BioImaging & Optics Platform (BIOP) Research Core Facility at the School of Life Sciences of EPFL and we are especially thankful for the assistance of Arne Seitz. We are grateful to Hugo Bellen, Benjamin White, Gerry Rubin, Vanessa Auld, Christopher Potter, Mel Feany, Serge Birman, Ed Kravitz, Chris Doe, and Pavel Tomancak for generating Drosophila stocks, software and providing advice. Stocks obtained from the Bloomington Drosophila Stock Center (NIH P40OD018537) were used in this study.

Senior Editor

  1. K VijayRaghavan, National Centre for Biological Sciences, Tata Institute of Fundamental Research, India

Reviewing Editor

  1. Sonia Sen, Tata Institute for Genetics and Society, India

Reviewer

  1. Sonia Sen, Tata Institute for Genetics and Society, India

Publication history

  1. Received: October 25, 2021
  2. Preprint posted: November 4, 2021 (view preprint)
  3. Accepted: June 9, 2022
  4. Version of Record published: July 8, 2022 (version 1)

Copyright

© 2022, Jiao et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 1,199
    Page views
  • 200
    Downloads
  • 1
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Wei Jiao
  2. Gard Spreemann
  3. Evelyne Ruchti
  4. Soumya Banerjee
  5. Samuel Vernon
  6. Ying Shi
  7. R Steven Stowers
  8. Kathryn Hess
  9. Brian D McCabe
(2022)
Intact Drosophila central nervous system cellular quantitation reveals sexual dimorphism
eLife 11:e74968.
https://doi.org/10.7554/eLife.74968
  1. Further reading

Further reading

    1. Cell Biology
    Amanda E Brandon, Lewin Small ... Gregory J Cooney
    Research Article Updated

    Obesity is generally associated with insulin resistance in liver and muscle and increased risk of developing type 2 diabetes, however there is a population of obese people that remain insulin sensitive. Similarly, recent work suggests that mice fed high carbohydrate diets can become obese without apparent glucose intolerance. To investigate this phenomenon further, we fed mice either a high fat (Hi-F) or high starch (Hi-ST) diet and measured adiposity, glucose tolerance, insulin sensitivity, and tissue lipids compared to control mice fed a standard laboratory chow. Both Hi-ST and Hi-F mice accumulated a similar amount of fat and tissue triglyceride compared to chow-fed mice. However, while Hi-F diet mice developed glucose intolerance as well as liver and muscle insulin resistance (assessed via euglycaemic/hyperinsulinaemic clamp), obese Hi-ST mice maintained glucose tolerance and insulin action similar to lean, chow-fed controls. This preservation of insulin action despite obesity in Hi-ST mice was associated with differences in de novo lipogenesis and levels of C22:0 ceramide in liver and C18:0 ceramide in muscle. This indicates that dietary manipulation can influence insulin action independently of the level of adiposity and that the presence of specific ceramide species correlates with these differences.

    1. Cell Biology
    Tricia T Nguyen, Gia K Voeltz
    Research Article

    Mitochondria are dynamic organelles that undergo cycles of fission and fusion at a unified platform defined by endoplasmic reticulum (ER)-mitochondria membrane contact sites (MCSs). These MCSs or nodes co-localize fission and fusion machinery. We set out to identify how ER-associated mitochondrial nodes can regulate both fission and fusion machinery assembly. We have used a promiscuous biotin ligase linked to the fusion machinery, Mfn1, and proteomics to identify an ER membrane protein, ABHD16A, as a major regulator of node formation. In the absence of ABHD16A, fission and fusion machineries fail to recruit to ER-associated mitochondrial nodes and fission and fusion rates are significantly reduced. ABHD16A contains an acyltransferase motif and an α/β hydrolase domain and point mutations in critical residues of these regions fail to rescue the formation of ER-associated mitochondrial hot spots. These data suggest a mechanism whereby ABHD16A functions by altering phospholipid composition at ER-mitochondria MCSs. Our data present the first example of an ER membrane protein that regulates the recruitment of both fission and fusion machineries to mitochondria.