(C) PLOS One
This story was originally published by PLOS One and is unaltered.
 . . . . . . . . . .



A developmental single-cell atlas of the Drosophila visual system glia reveals cell type diversification and subcellular mRNA compartmentalization [1]

['Amanda A. G. Ferreira', 'Department Of Biology', 'New York University', 'New York', 'United States Of America', 'Sergio Córdoba', 'Raghuvanshi Rajesh', 'Ben Jiwon Choi', 'Claude Desplan', 'Center For Genomics']

Date: 2026-10 

Glial cells are essential for proper nervous system development and function. To understand glial development and function, we comprehensively annotated the glial cells from two large single-cell mRNA-sequencing (scRNA-seq) atlases of the developing Drosophila visual system. This allowed us to identify all glial cell types from larval to adult stages and follow their developmental trajectories to understand how the diversity of glial types is generated during development. We show that whereas most glial types, such as chiasm glia, gradually change their transcriptome as they mature during development, neuropil glia that transcriptionally appear as a single-cell class in larvae, splits into ensheathing (EG) and astrocyte-like (ALG) glia types during pupal stages. We have experimentally validated these developmental trajectories and identified the genetic markers expressed through the differentiation between EG and ALG classes. Unexpectedly, our analysis of scRNA-seq datasets allowed us to discover that the transcriptome of glial cell bodies can be distinguished from that of their processes. We have identified that processes are enriched for distinct mRNAs that were validated in vivo. This work provides the most detailed transcriptomic analysis of optic lobe glia during development and helps explain the expansion of glial diversity observed in the adult visual system. We also present an innovative computational approach to identify mRNA species that are differentially localized to cell bodies or cellular processes.

Funding: This work was supported by the National Institutes of Health (R01EY017916 to C.D.; R01EY13010 to C.D.). S.C. was supported by an EMBO Long-Term Fellowship (ALTF 319-2019). A.A.G.F. and R.R. were supported by NYU’s GSAS MacCracken Program. A.A.G.F. was also supported by the Fleur Strand Graduate Fellowship (Biology Dept., NYU). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Data Availability: All raw and processed transcriptome data for the larval dataset are available at the GEO under accession number GSE167266. All raw and processed transcriptome data for the adult and pupa stages are available at the GEO under GSE142787. All processed datasets generated for the analyses in this manuscript are archived in Zenodo (DOI: https://doi.org/10.5281/zenodo.21878608 ). All related scripts used in this manuscript are available on GitHub https://github.com/AmandaAGF/Glia_Subcell_2023 . The code is archived in Zenodo (DOI: https://doi.org/10.5281/zenodo.21890456 ).

To comprehensively characterize the transcriptomic diversity of glia in the fly optic lobe and study glial development from larval to adult stages, we leveraged different scRNA-seq datasets at different stages of development that were previously generated in the laboratory [ 24 , 25 ]. About 32,000 sequenced optic lobe glial cells from larvae, four pupal stages, and adults were analyzed. We annotated most glial cells and validated these annotations at the third instar larval (L3) and adult stages. Their cell identities were then transferred to all pupal stages by integrating the datasets [ 24 ]. Unexpectedly, we have discovered that the mRNA expression levels of the classical glial marker repo are generally low and its expression alone is insufficient to identify all glial clusters in scRNA-seq datasets. We identified three new genes with broad expression among glial cell types (CG32032, AnxB9, and GstE12) that can be used along with repo to unambiguously identify all glial clusters. We also found that glial types specified in the larva follow three differentiation trajectories through adulthood: (1) glial types whose transcriptome minimally changed from larvae to adults, such as cortex and perineurial glia; (2) glial types whose transcriptome gradually changed and formed a linear trajectory, like inner and outer chiasm glia; and (3) glial types that formed an initial linear trajectory and then split into at least two different fates, such as medulla neuropil glia, which differentiates into astrocyte-like (ALG)/ensheathing glia (EG), and lamina neuropil glia that gives rise to epithelial/marginal glia in the adult. Moreover, while analyzing these scRNA-seq datasets, we discovered that certain glial cells were broken down during cell dissociation into different cell fragments that were then sequenced independently and clustered separately: unique glial cell types are thus represented by multiple clusters corresponding to the cell bodies and cell processes with distinct mRNA contents.

The Drosophila optic lobe has four main classes of glia: surface (perineurial and subperineurial), cortex, neuropil, and nerve-wrapping glia [ 8 , 15 – 17 ]. Surface glia surround the brain and form the blood-brain barrier that controls exchanges between the brain and the hemolymph [ 18 ]. Cortex glia is in close contact with the neuronal cell bodies, and its function is to nourish neurons and eliminate dead ones [ 7 ]. Neuropil glia is associated with neuronal terminals and is essential for synapse formation and neurotransmitter recycling [ 19 , 20 ]. Finally, nerve-wrapping glia wrap around axon bundles that cross from one neuropil to another, serving as support and protection for these neurons [ 15 , 17 , 21 ]. Glial cells are also involved in the development of the nervous system, contributing to axon pathfinding and neuronal circuit assembly [ 22 ]. Cortex glia influence the proliferation of neural stem cells (neuroblasts in flies) and the neuroepithelium to neuroblast transition [ 1 – 3 ]. Previous efforts to characterize glial cells in the Drosophila brain have relied on enhancer traps to identify different glial types based on their morphology [ 15 , 23 ]. More recently, the transcriptome of the glial cells of the adult optic lobe has been annotated [ 17 ]. However, the glial diversity in the larval brain and how it evolves through development until adulthood has not yet been investigated.

Anatomically, glia are highly polarized cells that have extensive cellular processes in which mRNAs are actively localized in specific subcellular compartments [ 11 – 13 ]. For example, the mRNAs encoding the major myelin protein MBP are found concentrated in the processes of glial cells surrounding the axon they ensheath, and their translation is triggered by the activity of the neuron [ 14 ].

Results

Identification of new glia-specific markers Marker genes are essential to identify glial clusters in scRNA-seq datasets. In Drosophila, glial cells are most often defined in vivo by the expression of the Repo protein [26]. To identify glial clusters in scRNA-seq, we selected the larval clusters [25] whose transcriptome exhibited a high correlation with the transcriptome of fluorescence-activated (FAC)-sorted cells based on the expression of a repo-GFP transcriptional reporter, and a low correlation with that of neurons FAC-sorted based on Elav expression, a pan-neuronal marker, as previously done for the adult optic lobe dataset [24] (Figs 1A and S1A). Surprisingly, the endogenous repo mRNA was lowly expressed and often absent in many of the glial clusters: repo mRNA was only significantly expressed in 11 out of 23 larval glial clusters and in 5 out of 21 adult glial clusters (Fig 1B and 1D). This low frequency of transcript detection in scRNA-seq experiments is probably due to ‘dropouts’ that are common for lowly expressed genes [27,28]. Consequently, the absence of repo detection in glial clusters in scRNA-seq data may result in many clusters mistakenly not being considered glial cells. For example, in the larva both types of surface glia (subperineurial, spg and perineurial, pg) had low repo expression levels in the dataset (cluster 21 for spg, and 8 and 14 for pg, Fig 1B), which correlates to the low detection of repo mRNA by Hybridization Chain Reaction RNA fluorescence in situ hybridization (HCR RNA FISH) [29,30] even if these cells could easily be identified as glial cells by antibody detection of the Repo protein (S1D Fig). PNG larger image

TIFF original image Download: Fig 1. Identification of new glia-specific marker genes. (A) Schematic representing how glial cell clusters were defined. Pearson correlation between the transcriptome of the clusters with that of bulk transcriptome of repo-positive and elav-positive cells was used to define glial clusters. The defined clusters were used to find new glial markers. (B) Dot plot showing the newly identified glial markers in the larva (L3). Numbers identify individual glial clusters, and “Non-Glial Cells” represent the combined gene expression of all the remaining cells from the whole larva dataset from Konstantinides and colleagues, 2022. The color of the dots indicates the mean normalized gene expression across all cells in a group (Average Expression), while the dot size indicates the fraction of cells with nonzero expression (Percent Expressed). (C) Schematic representation of the L3 optic lobe, indicating the position of the different glial cell types (in pink) and the localization of the neuropils (dark gray). The blue section represents the plane imaged in the immunostaining shown in (F-F’‘‘). (D) Dot plot showing the newly identified glial markers in the adult dataset per glial cluster. “Non-Glial Cells” represent the combined gene expression of all the remaining cells from the whole adult stage dataset from Özel and colleagues, 2021. (E) Schematic representation of the adult optic lobe, indicating the position of the different glial cell types (in pink) and the localization of the neuropils (dark gray). (F-F’‘‘) Immunostaining of the larval (L3) optic lobe showing the localization of Repo, CG32032 and AnxB9 antibodies. (G-G’‘‘) Immunostaining of the adult optic lobe showing the localization of Repo, CG32032 and AnxB9 antibodies. CB, Central Brain, fg = fenestrated glia, pg = perineurial glia, spg = subperineurial glia, dsg = distal satellite glia, psg = proximal satellite glia, eg = epithelial glia, mg = marginal glia, xgo = outer chiasm glia, xgi = inner chiasm glia, cxg = cortex glia, mng = medulla neuropil glia, rwg = retina wrapping glia. Scale bar = 20 µm. The code used to generate the graphs can be found in https://zenodo.org/records/21890456. https://doi.org/10.1371/journal.pbio.3004015.g001 To overcome this challenge, we looked for other glia-specific genes with higher expression that could allow us to robustly identify glia in scRNA-seq datasets, independently or together with repo expression. By comparing all the larval glial clusters with the other nonglial clusters, we identified several genes that were differentially expressed in glial cells as compared to neurons: Ama, MtnA, AnxB9, CG32032, CG7433, GstE12, CG9686, CG15209, and Npc2a (S1B and S1C Fig). Among these genes, AnxB9, CG32032, and GstE12 were individually expressed in most glial clusters, including those that did not express repo (Figs 1B, 1D and S1C) but were largely absent in neuronal clusters, with few exceptions. AnxB9 belongs to the family of annexins that bind to phospholipids and participate in membrane remodeling, membrane trafficking and exocytosis [31]. Annexins are widely found in the nervous system and show different expression patterns in neurons and glia [32,33]. CG32032 is predicted to belong to the calycin protein superfamily (FlyBase [34], which includes lipocalins and fatty acid-binding proteins often found in glial cells, such as the human FABP7 and PMP2 [6,35]. Finally, GstE12 encodes a cytoplasmic Gluthatione S-Transferase (GST) enzyme [36]. In mice, astrocytes enhance proinflammatory response in the brain through GSTs, and the knockdown of these enzymes in astrocytes compromises microglial activation [7]. We generated antibodies for AnxB9 and CG32032, both of which showed highly glia-specific signal in immunostainings, which confirmed that the protein products of CG32032 and AnxB9 genes are present in glial cells in both L3 larvae (Fig 1F and schematic representation in C) and adults (Fig 1G and schematic representation in E). However, we observed some cell-type specific variability: in L3 brains AnxB9 staining was high in giant chiasma glia (xgo) and low in epithelial and marginal glia (eg/mg), while CG32032 antibody signal was low in xgo and high in eg/mg (S1E and S1F Fig). Thus, CG32032 and AnxB9 are both expressed strongly and mostly specifically in glia and, in combination with repo, are excellent candidates for the identification of glial cell clusters in scRNA-seq datasets. Based on the expression of at least two of these genes, all previously identified glial clusters in both larval (23 out of 23 clusters) and adult (21 out of 21 clusters) datasets were identified, and none of the nonglial clusters expressed these combinations. To test whether these genes could help identify glial types in other datasets, we analyzed their expression in a published optic lobe scRNA-seq dataset from pharate adults (96 hours pupation – P96) [37], and found that repo was lowly expressed in three clusters annotated as glia by the authors (S1G Fig). However, these three clusters expressed at least two of the new glial markers. Moreover, this dataset contained clusters called “C clusters” that were annotated neither as glia nor as neurons. Many of the cells in these “C clusters” expressed the new glial markers. Indeed, out of all 321 cells in the “C clusters”, 103 had transcriptomes that resembled adult subperineurial glial cells. Similarly, cluster N183, previously annotated as an unknown neuronal cluster, also showed expression of these glial markers and was composed of 19 cells, 9 of which also correlated with subperineurial glial cells (S1G Fig). Another published Drosophila scRNA-seq dataset of the adult ventral nerve cord (VNC) was also analyzed [38]. Two of the clusters that the authors had identified as glia/neuron mix due to low expression of repo could be confidently identified as glia when using the combined expression of repo, CG32032, AnxB9, and GstE12 (S1H Fig, arrows) and indeed the authors later managed to identify these clusters as glia, based on specific glial subtypes markers [39]. Interestingly, in this VNC dataset a number of clusters that did not express any of the other glial markers did exhibit AnxB9 expression. Accordingly, we observed strong AnxB9 antibody signal in nonglial cells, presumably neurons, in the larval VNC (S1I Fig). In summary, using four different datasets, we showed that repo, CG32032, AnxB9, and GstE12 are marker genes that can be used in combination to consistently identify glial clusters in Drosophila scRNA-seq datasets.

Annotation of glial clusters in a developmental scRNA-seq atlas of the fly optic lobe Next, we sought to evaluate glial diversity in the larval and adult stages. High-resolution transcriptomic atlases of the optic lobe throughout development [24,25] had identified more than 200 cell types, including progenitors, neurons and glia, but glial clusters were not annotated as specific glial types. We excluded neurons and re-clustered the bona-fide glial cells (Fig 2B) [40]. Glial cell bodies occupy stereotypical positions in the brain and consistently express Repo protein, which allowed us to match specific gene expression from each cluster to known glial cell types. We annotated and validated larval clusters based on known [7,15,20,23,39,41–48] or new markers identified in this study (summarized in S1 Table and Methods) using transcriptional reporters, protein traps, enhancer traps, antibodies, as well as specific HCR RNA FISH probes (Fig 2A and 2C–2M). PNG larger image

TIFF original image Download: Fig 2. Cluster validations for larva glial annotations. (A) Dot plot showing the expression levels of the genes used for the validation of each annotated cluster in L3. (B) Left, hierarchical tree showing cluster annotations after validations reveals 10 cell types (in bold text), excluding cells of central brain origin. Note that Chiasm glia was manually separated into Inner and Outer Chiasm subtypes. Right, unsupervised clustering of larval glial cells. (C–M) Immunostaining and FISH analysis of marker gene expression in the L3 optic lobe. Neuropil structure is revealed by N-cadherin (Ncad) or HRP staining (yellow), while Repo (magenta) is used to label glia nuclei in the left panels. Marker gene expression is shown using antibodies, lacZ reporters, protein tags or specific Gal4 lines driving UAS-nGFP expression (cyan). (C) rost-lacZ reporter (yellow) is only expressed in subperineurial glia. Anti-Svp antibody (cyan) labels both perineurial and subperineurial glia. (D) Optix (cyan) is present in satellite, epithelial, and marginal glia but not in subperineurial, outer chiasm, and medulla neuropil glia. (E) Dll (cyan) labels epithelial, marginal, lobula complex neuropil and medulla neuropil glia but not satellite, outer, and inner chiasm glia. (F) DAT-Gal4 (cyan) is expressed in the outer and inner chiasm glia. It is less frequently observed in the inner chiasm glia. (G) Hth (cyan) antibody is higher in inner chiasm than in outer chiasm glia. (H) Hand-Gal4 (cyan) is only expressed in medulla neuropil glia. (I) HCR RNA FISH probes against Cpr51A (yellow) and repo (magenta) mRNA. Hoechst antibody is shown in gray. Cpr51A is expressed in the chiasm and lobula complex glia. The image shows the standard deviation projection of three consecutive focal planes 2 µm apart. (J) GFP-tagged CG4288 (cyan) is expressed only in the cortex and lobula complex glia but not in the perineurial, medulla neuropil, and inner chiasm glia. Asterisk indicates Bolwig organ. (K) GFP-tagged alphaTub85E (cyan) is expressed in retina wrapping glia. (L) GFP-tagged Npc2b (cyan) is expressed only in cortex glia from the central brain. (M) Adgf-D (cyan) is expressed in the central brain glia. pg = perineurial glia, spg = subperineurial glia, sg = satellite glia, eg/mg = epithelial and marginal glia, mng = medulla neuropil glia, xgo = outer chiasm glia, xgi = inner chiasm glia, cxg = cortex glia, rwg = retina wrapping glia, cb-cxg = central brain cortex glia, lcng = lobula complex neuropil glia. Gal4-UAS (>>), GFP/LacZ from construct (::). Scale bar = 20 µm. The code used to generate the graphs can be found in https://zenodo.org/records/21890456. https://doi.org/10.1371/journal.pbio.3004015.g002 At the adult stage, we based our annotations on cell-type specific markers established by earlier studies that had identified genes expressed in specific subtypes of glia using enhancer traps, antibody staining, cell-type specific RNA-seq, and scRNA-seq [20,39,46,49–54]. We determined the identity of adult clusters by taking advantage of the spatial distribution of glial cells combined with immunohistochemistry for genes that were expressed in one or a few clusters (Fig 3B and 3C–3L; see S2 Table for the list of markers, annotations, and corresponding references). Ultimately, we could assess the transcriptome of all glial cell types and determine specific markers for each of them (S3 Table). This represents a powerful resource for studying glial cells and their contribution to the development and function of the optic lobe. PNG larger image

TIFF original image Download: Fig 3. Cluster validations for adult glial annotations. (A) Dot plot showing the expression levels of the genes used for the validation of each annotated cluster in adult. (B) Left, hierarchical tree showing cluster annotations after validations reveals 14 distinct cell types (in bold text) plus a number of clusters annotated preliminarily as Astrocyte- and Ensheathing-like glia. Right, unsupervised clustering of adult glial cells. (C–L) Immunostaining analysis of adult stage optic lobe. Neuropil structure is revealed by Ncad or HRP staining (yellow), and Repo (magenta) labels glia nuclei in the left panels. Marker gene expression is shown using antibodies, lacZ reporters, protein tags, or specific Gal4 lines (cyan). (C) Tret1-1-Gal4 (cyan) is only expressed in perineurial glia and not in fenestrated. (D) Vmat-Gal4 (cyan) is expressed in both perineurial and fenestrated glia. (E) rost-lacZ reporter (cyan) is expressed only in subperineurial glia. (F) GFP-tagged CG3036 (green) is expressed only in chalice glia. Vvl antibody (cyan) labels distal and proximal satellite glia. Optix (yellow) is present in chalice, marginal, epithelial, and proximal satellite glia. (G) b-Gal4 (cyan) is expressed in distal and proximal satellite glia and in epithelial glia, but not in marginal. (H) GstT4-Gal4 (cyan) is expressed only in marginal glia. (I) DAT-Gal4 (cyan) is expressed in outer chiasm glia but only seen rarely in inner chiasm glia. (J) GFP-tagged CG31663 (cyan) is expressed in proximal satellite glia and some medulla neuropil glia. This might reflect ensheathing- (mng+) vs. astrocyte-like (mng−) identity. It is also expressed in outer and inner chiasm glia and in cortex glia. (K) ome-Gal4 (cyan) is expressed in inner chiasm glia, marginal glia and some medulla neuropil glia, which could reflect ensheathing (mng+) vs. astrocytes-like (mng−) identity. (L) nkd-lacZ reporter (cyan) is expressed in medulla neuropil glia but not in lobula neuropil glia. fg = fenestrated glia, pg = perineurial glia, spg = subperineurial glia, dsg = distal satellite glia, psg = proximal satellite glia, chg = chalice glia, eg = epithelial glia, mg = marginal glia, xgo = outer chiasm glia, cxg = cortex glia, lng = lobula neuropil glia, mng = medulla neuropil glia. Gal4-UAS (>>), GFP/LacZ from construct (::). Scale bar = 20 µm. The code used to generate the graphs can be found at https://zenodo.org/records/21890456. https://doi.org/10.1371/journal.pbio.3004015.g003

Developmental trajectories of glial cells from larva to adult stages In the course of annotating glial types in larva and adult stages, we noticed that the genetic markers of glial types changed over time, and that there was a significant increase in glial cell diversity in the optic lobe, from 10 cell types in larva to at least 14 distinct cell types in adult (Figs 2B and 3B). To investigate this increase, we integrated all glial cells from larvae [25], P15 (15 hours after pupation), P30, P50, P70, and adults [24] into one dataset (Fig 4A). Interestingly, the same cell types across stages clustered together, such as cortex, perineurial, subperineurial, lamina neuropil, medulla neuropil, inner and outer chiasm glia (S2A and S2B Fig). The transcriptomic similarity of the same cell types throughout development corroborates the robustness of our marker-based cell type annotation (S1 and S2 Tables). PNG larger image

TIFF original image Download: Fig 4. Developmental diversification of neuropil glia. (A) Left, UMAPs representing all glial clusters from L3, P15, P30, P50, P70 and adult stage optic lobes. Right, UMAP integration of all developmental stages; cells are color-coded according to their stage of origin. Blue arrows highlight glial cells that show a gradual developmental trajectory from L3 to adult in our integration. Red arrows indicate trajectories where glial cells that are initially similar bifurcate into different terminal fates. Circled in black are glial cells that do not show a developmental trajectory from L3 to adult according to the integration. Dashed lines represent trajectories that do not contain cells from larval stage. (B) UMAP showing isolated medulla neuropil glia from the integrated dataset in (A), where the divergent trajectories of medulla astrocyte-like (ALG) and can be observed. Hand displays a salt-and-pepper expression that is restricted at the beginning of the trajectory (L3). (C–E) L3 expression of 10C12-Gal4, 54H11-Gal5, and Hand-T2A-Gal4 lines, driving the UAS-nlsGFP-PEST (C and E) or UAS-LifeActGFP (D) in cyan. Ncad (yellow) is used to visualize the neuropil and Repo (magenta) to label all glia. White dashed boxes delineate the area magnified in the panels below. 10C12-Gal4 is expressed specifically in epithelial and marginal glia, as well as in Lamina wide field neurons. 54H11-Gal4 labels epithelial and marginal glia and medulla neuropil glia. Hand-T2A-Gal4 expression is only observed in some medulla neuropil glia (filled arrowheads), and absent in others (empty arrowheads). (F–H) FLEXAMP memory cassette is used to immortalize L3 expression of the Gal4 lines in C–E and to observe the morphology of adult glia. Only UAS-myrGFP expression is shown, and the neuropil shape is delineated in yellow dashed lines. Lamina and close-ups of medulla neuropil are indicated. Note that epithelial and marginal glia, that are indistinguishable at L3, diverge into specific morphologies and localizations in the adult (F and G). Tracings of L3 medulla neuropil glia show that they give rise to all morphotypes (ensheathing and astrocyte-like) and localizations (proximal, lateral and distal) of adult medulla neuropil glia (G and H). xgo = outer chiasm glia, mng = medulla neuropil glia, lng = lamina neuropil glia, eg = epithelial glia, mg = marginal glia. Gal4-UAS (>>). Scale bar = 20 µm. The code used to generate the graphs can be found at https://zenodo.org/records/21890456. https://doi.org/10.1371/journal.pbio.3004015.g004 We observed three distinct types of developmental trajectories of glial cell types based on our integrated dataset: (1) Cell types that did not exhibit a trajectory (i.e., minimally changed) from larva to adult, such as cortex and perineurial glia. These are cells that remain transcriptionally stable throughout development (Fig 4A, black circles). (2) Cell types that gradually changed their transcriptome in a linear trajectory throughout developmental stages, like inner and outer chiasm glia, likely reflecting a maturation process (Fig 4A, blue arrows). We also observed certain discrepancies between larval and adult cell types, which can be explained by previous literature. For example, in adults there are two satellite glia in the lamina named proximal and distal [16]. However, only one satellite glia is found in larvae, the larval satellite glia that gradually matures to become proximal satellite glia in adults. In contrast, the adult distal satellite glia has a different origin, corresponding to larval retina wrapping glia [23], which is a larval nerve-wrapping glial type that is repurposed to work as cortex glia in adults. (S2A and S2B Fig). (3) Cell types that formed an initial linear trajectory and then split into at least two different cell fates (Fig 4A, red arrows). Notably, all glial types with a divergent trajectory were neuropil-associated glia that split from a common larval identity to adult ALG and EG cell types.

Specification and diversification of lamina and medulla neuropil glia In the adult lamina neuropil, there are two types of neuropil glia, known as epithelial glia (eg, a type of ALG) and marginal glia (mg, a type of EG). eg/mg originate in the larva from specialized precursors localized in a region of the OPC called glia precursor center (GPC) [45]. At this stageeg and mg cannot be separated transcriptionally into distinct cell types, but later split into separated trajectories (Figs 4A, S2A and S2B). We further investigated this split by UMAP trajectory analysis (S3A Fig) and found that eg and mg transcriptomes diverge soon after they are born, i.e., between L3 and P15 (S3B and S3C Fig). This apparent split into distinct cell types early during metamorphosis would explain the expanded cell diversity in the adult as compared to the larval scRNA-seq annotations. To confirm this trajectory, we performed memory-tracing experiments using the FLEXAMP memory cassette [55], inducing recombination at late L3 using 10C12-Gal4, a driver expressed in epithelial and marginal glia (ege.g.,/mg) as well as in Lamina wide-field (Lawf) neurons [45] (Fig 4C). In the adult brain, we recovered both epithelial (eg) and marginal glia (mg), confirming that lamina neuropil glia diverge into distinct cell types during pupal development (Fig 4F). This result was independently confirmed using FLEXAMP with the 54H11-Gal4 driver, which is expressed in eg/mg and medulla neuropil glia (see below). The adult medulla neuropil presents two different neuropil glia subtypes, medulla EG (which have long projections closely associated with photoreceptors that show limited branching) and medulla ALG (that send dense, highly branched fine projections) [16]. Both EG and ALG can present diverse morphologies and positions around the neuropil, defined previously as ‘morphotypes’ [16,17]. Both types derive from larval medulla neuropil glia (mng), which originate from the last division of neuroblasts in the outer proliferation center (OPC) [56,57] and present a single trajectory that remains transcriptionally homogeneous from larva to P30 and then splits into two different trajectories leading to the EG and ALG fates (Figs 4A, 4B, S2A and S2B). To validate this predicted trajectory, we induced FLEXAMP recombination in L3 using the 54H11-Gal4 driver, which is expressed in medulla neuropil glia (mng) and eg/mg glia in larva (Fig 4D). Besides eg/mg lamina glia, we were able to recover adult medulla neuropil glial cells corresponding to both expected subtypes, EG and ALG glia. Furthermore, all adult morphotypes were observed, including distal and lateral medulla EG, as well as distal, lateral, and proximal medulla ALG [16,17] (Fig 4G). To further validate this result, we used another driver, Hand-T2A-Gal4. Although this driver is highly specific for L3 mng, not all larval mng were labeled by GFP (Fig 4E). This sparsity was corroborated by our single-cell analysis, where only 26% of L3 mng expressed Hand (Fig 4B). We used a memory-lacZ cassette [58] to trace L3 mng expressing Hand-T2A-Gal4 in the adult and found that Hand expressing cells were all Repo+ and localized around the medulla neuropil, consistent with the position of adult medulla neuropil glia (S2C Fig). We used FLEXAMP to assess their morphology and were able to identify both medulla EG and ALG cells of all morphotypes [16,17]. These results confirm that L3 mng initially share a common transcriptional identity and diverge into distinct EG and ALG subtypes during pupal development. These findings in lamina and medulla glia support the inferred bifurcation in the single-cell trajectory analysis.

Molecular analysis of medulla neuropil ALG and EG diversification To further analyze the basis of this transcriptomic divergence, we examined the trajectory of mng which splits into the medulla ALG and EG subtypes during pupal development. To identify the exact developmental stage of the split, we manually subset the cells at the fork of the trajectory and found that these cells were mostly coming from the P30 dataset (Fig 5C, inset), suggesting that transcriptomic divergence of mng occurs much after their birth, unlike the lamina glia that split shortly after L3 (compare Fig 5A–5D with S3A–S3C Fig). We then embedded a pseudotime trajectory using Monocle3 [59] on the existing UMAP reduction, which allowed us to closely recapitulate the developmental stages of the dataset as a function of pseudotime (Fig 5B, 5D, and see Methods). Since TFs are the main drivers of cell fate, we focused on their expression to explore ALG and EG fate divergence. We found 63 TFs that were differentially and significantly modulated along the pseudotime trajectory (Fig 5E), and we then used K-means to cluster TFs enriched at early, mid, and late pseudotime points (see Methods). We found that TFs known to be required for mng specification, such as gcm and gcm2, as well as Hand, were expressed early in development, as expected [42,57]. Interestingly, we also observed early expression of canonical Notch effector genes such as Hey and five E(spl) complex TFs [60,61], strongly suggesting that mng have active Notch signaling early after specification. This NotchON status and initial Hey expression is also shared by lamina neuropil glia [45]. We validated the expression of Hey in eg/mg and mng by immunostaining, which showed that the presence of Hey protein in nascent lamina and medulla neuropil glia is transient and disappears soon after mng migration (Fig 5F and 5F’). K-means clustering of TFs along pseudotime of the entire trajectory revealed that four genes, crc, Eip78C, gem and CG3328 were enriched at the bifurcation point of ALG and EG subtypes, providing a list of candidate TFs that might be involved in mng subtype specification. Furthermore, we expanded this candidate TF list by analyzing the ALG and EG trajectories separately (S3D, S3E and S3D’, S3E’ Fig, respectively), which will serve as the basis for future perturbation-based experiments. Overall, our analysis provides a transcriptomic basis for mng subtype specification during optic lobe development. PNG larger image

TIFF original image Download: Fig 5. Transcriptomic divergence of medulla neuropil glia along development. (A) UMAP of medulla neuropil glia (mng) cells showing the divergence of astrocyte-like (ALG) and ensheathing (EG) glial cell types. The terminal fates identified in the adult dataset are circled. (B) UMAP showing cells colored by their developmental stages of origin. (C) UMAP-based subsetting of cells at the fork (colored red) between ALG and EG adult fates. Inset displays the proportion of red cells that belong to each developmental stage. Note that the most represented stage at the trajectory fork is P30. (D) UMAP showing pseudotime alignment of mng glial cells. This pseudotime embedding is used for analyzing gene expression dynamics in (E). (E) Heatmap of dynamically regulated TFs along pseudotime highlighting putative TFs associated with mng divergence (dashed box) and canonical Notch effector genes (blue arrowheads). (F, F’) L3 brain showing the lamina (F) and medulla (F’) neuropil glia stained for Hey (magenta) and Repo (cyan) antibodies. Both lamina and medulla neuropil glia express Hey after specification (full arrowheads) and turn it off shortly after (empty arrowheads). eg = epithelial glia, mg = marginal glia and mng = medulla neuropil glia. The code used to generate the graphs can be found at https://zenodo.org/records/21890456. https://doi.org/10.1371/journal.pbio.3004015.g005

Specification of lobula neuropil glia The glia associated with the lobula complex neuropil (which during pupal development forms the lobula and lobula plate neuropils) seems to develop in a more complex manner than what we found for medulla and lamina neuropil glia. We annotated and validated one cluster as lobula complex neuropil glia in the larva (Fig 2A, 1B and 1I). However, unique markers to validate lobula neuropil glia in adults were not found (Fig 3A). Since we confidently annotated all remaining glial types, the 6 unannotated glial clusters in the adult were good candidates to be lobula neuropil glia. Moreover, these unknown adult clusters were closely associated with the L3 lobula complex neuropil glia cells in our age-integrated dataset (Fig 4A). To gain further insight into the presumptive lobula neuropil glia, we integrated these clusters across developmental stages (S2E and S2F Fig). We named the presumptive ALG clusters “Astrocytes_1/ 2/ 3/ 4” and presumptive EG clusters “Ensheating_1/ 2/” (S2F Fig). We observed two parallel trajectories: one for lobula neuropil ALG, which originated in the validated L3 lobula complex neuropil glia cluster, and another for lobula neuropil EG, that started at P15 (S2F Fig). Interestingly, the L3 lobula complex cluster already expressed mature ALG markers and was positioned between adult medulla ALG and the unknown adult astrocyte-like clusters in the integrated data UMAP (see “Lobula C. Neuropil” and “Astrocytes_1/ 2/ 3/ 4”, S2A and S2B Fig). This suggests that the lobula complex neuropil glia found in the larva are already mature, astrocyte-like cells. Intriguingly, the lobula complex is physically closely associated with the central brain during larval development [19]. In the central brain, adult (or secondary) neuropil glia are produced during larval stages from type II neuroblasts to replace larval (or primary) neuropil glia born during embryogenesis, raising the possibility that the L3 lobula complex neuropil glia cluster was composed of mature astrocyte glia migrated from the central brain. Accordingly, we have observed that a population of glial cells surrounding the lobula complex neuropil that expressed 54H11-Gal4 appeared to migrate from the central brain (S2D Fig). It is possible that these cells account for the lobula complex neuropil cluster of larval origin that is at the base of the lobula neuropil ALG trajectory (S2E Fig, yellow cells). Meanwhile, the presumptive lobula neuropil EG trajectory, which gives rise to “Ensheathing_1/ 2/” adult clusters, started at P15 and did not include any L3 cells (S2E, S2F Fig). We looked for genes that were expressed early in both medulla and lamina neuropil glia developmental trajectories, and found 61 common genes (S4 Table), six of which were unique to these clusters and mostly absent in other glial types: CR30009, gcm, E(spl)mgamma-HLH, CG11670, E(spl)malpha-BFM, and Traf4 (S4A–S4F Fig). Interestingly, these genes were also expressed at the beginning of the lobula neuropil EG trajectory but were almost absent from the lobula neuropil ALG trajectory (S4G–S4I Fig). This observation supports the idea that ALG lobula neuropil glia are more mature than the rest of neuropil glia at L3 and suggests that EG lobula neuropil glia are born at later stages (P15). Finally, there were three clusters of surface lamina glia for which we could not identify contributions from larva glia: pseudocartridge (a subtype of subperineurial glia) as well as fenestrated and chalice glia, which are both subtypes of perineurial glia (Figs 4A, S2A and S2B, dashed circles) [16]. Pseudocartridge, fenestrated glia and likely chalice glia originate from the eye disc [62] that was not sequenced in the larva dataset [25], thus explaining their absence (summarized on S5 Table). Together, our data show that the increased number of adult glial clusters compared to the larval dataset is, in part, the result of increased glial cell diversity originated from larval neuropil glia that diverges into different subtypes during pupal development.

[END]
---
[1] Url: https://journals.plos.org/plosbiology/article?id=10.1371/journal.pbio.3004015

Published and (C) by PLOS One
Content appears here under this condition or license: Creative Commons - Attribution BY 4.0.

via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/plosone/