Concurrent temporal patterning of neural stem cells in the fly visual system

Abstract

In the Drosophila optic lobe, medulla neuroblasts (NBs) are patterned by temporal and spatial inputs, which contribute to the generation of the medulla’s ~150 neuron types. Here, we describe a third patterning mechanism that further diversifies neuronal fates in the medulla. The neuroepithelium from which NBs are continuously produced is patterned by opposing temporal gradients of the Imp and Syp RNA-binding proteins and their downstream transcription factors. Imp/Syp patterning results in the generation of seven cell types in successive developmental windows from NBs at the Vsx1–Hth spatial–temporal address. Medulla NBs are thus patterned by two concurrent temporal mechanisms: (1) the Imp/Syp state of the neuroepithelium when they are generated; and (2) the transcription factor cascade that progresses within them as they age. We further find that the birth order of medulla neurons correlates with their position along the anterior-posterior axis of the adult cortex, resulting in unanticipated regionalization of the retinotopic map.

Introduction

The development of complex neural circuits requires that stem cells generate diverse types of neurons in the correct number and location. Pioneering work in insects and vertebrates has shown that the spatial and temporal patterning of neural stem cells plays a critical role in regulating multiple aspects of neurogenesis^1,2. In the Drosophila embryonic ventral nerve cord, neural stem cells, termed neuroblasts (NBs), integrate spatial and temporal inputs at both the transcriptional and epigenetic levels to generate neural diversity^3,4,5,6. In vertebrates, temporal and spatial patterning has been observed in multiple brain regions, including the neural tube, where localized morphogen activity assigns unique positional identities to neural stem cells and distinct sets of transcription factors (TFs) are expressed in neurons based on their developmental birth order^{7,8,9,10,11,12,13}.

In recent years, the Drosophila medulla has emerged as a powerful model system in which to study how neural stem cells integrate spatial and temporal patterning inputs to generate neuronal diversity. The medulla, together with the lamina and lobula complex, comprises the optic lobe and mediates the processing of both color and motion information^14,15,16,17. Its 40,000 cells are organized into ~800 repeating columns that propagate the retinotopic inputs received from the overlying ommatidia^17,18. The ~150 neuronal cell types of the medulla develop from a neuroepithelial (NE) crescent termed the outer proliferation center (OPC)^{19,20,21,22,23,24,25}. Beginning at the onset of the third larval instar stage, and continuing for 3 days, a proneural wave moves in a medial-to-lateral direction to convert NE cells into NBs^20,26,27,28. In the wake of the proneural wave, the youngest NBs are located laterally, closest to the OPC NE, whereas the oldest NBs are found medially, adjacent to the central brain^20,27,28. Once specified, NBs undergo multiple asymmetric divisions to generate ganglion mother cells (GMCs), which then divide once more to produce the neurons and glia of the medulla^19,26.

The integration of spatial and temporal inputs contributes to the generation of neuronal diversity in the medulla^23,29. In the spatial axis, the OPC NE from which the NBs are derived is patterned by the expression of three homeobox TFs: Visual system homeobox 1 (Vsx1) in the center of the crescent, Optix in the arms, and Retinal homeobox (Rx) at the tips²³. The Rx region is further subdivided by the expression of the signaling molecules, Decapentaplegic (Dpp) and Wingless, while the mOPC is divided into three sub-domains by the expression of Dpp and the transcriptional repressor, Brinker^30,31. Lastly, two zinc finger TFs, Spalt and Disco, compartmentalize the crescent into dorsal and ventral halves³². In the temporal axis, a cascade of temporal TFs (tTFs)—including the sequentially expressed Homothorax (Hth), Eyeless (Ey), Sloppy-paired 1/2 (Slp1/2), Dichaete (D), and Tailless (Tll) proteins—patterns NBs as they age, resulting in the production of at least 11 unique temporal NB windows^21,33,34. Distinct neuronal cell types are generated by NBs based on their spatial–temporal address²³. For example, Pm3 neurons are generated by NBs derived from the Vsx1 spatial compartment and Hth temporal window²³.

Medulla NBs generate two types of neurons: uni-columnar neurons that are made in all OPC spatial compartments and multi-columnar neurons that are compartment-specific in their origin²³. Uni-columnar neurons are generated independent of spatial patterning, relying solely on temporal inputs for their specification. These neurons are made in large numbers and do not migrate during development, remaining where they are born through to the adult, with each contributing to one of the 800 columnar units that are distributed along the dorsal–ventral (D–V) and anterior–posterior (A–P) axes of the medulla. In contrast, multi-columnar neurons, which each contribute to multiple medulla columns and are thus produced in fewer numbers, require the integration of both spatial and temporal inputs for their specification. These neurons are generated in spatially restricted regions along the D–V axis and subsequently migrate along the D–V axis during pupal development to occupy all regions of the adult medulla. While significant progress has been made in understanding the mechanisms of cell-type specification and reorganization along the D–V axis, it is not known whether progenitor patterning contributes to the specification and distribution of neuron types along the A–P axis.

Here, we describe a third patterning axis that operates together with the previously reported spatial and temporal axes to significantly increase neuronal diversity in the medulla. We demonstrate that opposing gradients of the RNA-binding proteins, IGF-II mRNA-binding protein (Imp) and Syncrip (Syp), temporally pattern the NE from which the NBs are derived. Imp and Syp regulate the downstream TFs Chronologically inappropriate morphogenesis (Chinmo), Maternal gene required for meiosis (Mamo) and Ecdysone-induced protein 93F (E93) within the NE to assign NBs unique temporal identities based on when during development they are generated by the proneural wave. Imp and Syp have been previously shown to act in central brain NBs to generate neuronal diversity^35,36,37. Our finding that Imp and Syp temporally pattern the NE allows for medulla NBs to be concurrently patterned by two independent temporal mechanisms: (1) the Imp/Syp state of the NE from which each NB is generated and (2) the tTF cascade that progresses within all NBs as they age. As a proof of principle, we show that Imp/Syp patterning diversifies the progeny of Vsx1–Hth NBs; in addition to the previously reported Pm3 neurons, six additional multi-columnar cell types are generated at this spatial–temporal address in an Imp and Syp-dependent manner. Based on the expression of the Imp/Syp gradients in all OPC spatial compartments and neuroblast temporal windows, the diversification of neuronal fates by Imp/Syp likely extends to all OPC spatial–temporal addresses. We further show that, in contrast to the extensive migration observed for multi-columnar neurons along the D–V axis, Vsx1–Hth-derived neurons occupy stereotyped positions along the A–P axis of the adult medulla in accordance with their developmental birth order; earlier-born neurons are located anteriorly, whereas later-born neurons are located posteriorly. Despite their A–P cell body localization, the majority of these neurons possess arborizations that innervate columns throughout the A–P axis of the medulla. However, we find that three cell types, TmY12, TmY17 and Tm26, only innervate the medulla columns located closest to their cell bodies, which results in unanticipated patterning of the fly’s retinotopic map along the A–P axis. The concurrent temporal patterning of symmetrically and asymmetrically dividing neural stem cells thus acts as a powerful mechanism by which to couple neurogenesis with circuit assembly.

Results

Unexpected neuronal diversity at the intersection of the Vsx1 spatial and Hth temporal axes

Previous work has shown that Pm3 neurons are born at the intersection of the Vsx1 spatial and Hth temporal axes and co-express these two TFs throughout their development²³ (Fig. 1a). To allow for the visualization and genetic manipulation of developing Pm3 neurons, we generated a split-Gal4 line³⁸ comprised of vsx1-Gal4DBD and hth-VP16AD CRISPR-mediated insertions (Fig. 1a). In the larval, pupal and adult optic lobe, vsx1⋂hth-Gal4 > GFP expression is restricted to Vsx1 and Hth co-expressing cells, indicating that the line is a faithful reporter for neurons born at this spatial–temporal address (Supplementary Fig. 1a–c). Surprisingly however, we observed that the neurons labeled by this line send projections to multiple layers in the adult medulla and lobula complex, exhibiting an arborization pattern that is more widespread than that predicted for Pm3 neurons alone, which arborize exclusively in layers M8–M10 of the medulla (Fig. 1b). This unexpected arborization pattern suggests that, along with the previously reported Pm3 neurons, additional neuron types are generated at the Vsx1–Hth spatial–temporal address.

Fig. 1: Seven distinct neuron types are labeled by the vsx1∩hth-Gal4 driver.

a Schematic of the split-Gal4 system used to drive reporter expression in Vsx1 and Hth co-expressing neurons (green). b Visualization of vsx1∩hth-Gal4 > GFP (green) in the adult optic lobe. The medulla (Me), lobula (Lo), and lobula plate (Lp) are indicated with white dashed lines. Pm3 neuron arborizations are visible in the proximal medulla layers (white arrow) in addition to arborizations of other cell types (yellow arrowheads). c–h MCFO clones (V5/HA, green) and illustrations of the six morphologically distinct neuron types labeled by vsx1∩hth-Gal4 in the adult optic lobe. For each neuron, 3D reconstructions of MCFO clones are on the left, and illustrative renderings of their projections across the optic lobe neuropils are on the right. The M7 medulla layer is shaded in gray, and yellow arrowheads denote the neuronal cell body. i tSNE visualization of vsx1 and hth co-expressing cells subsetted from a whole adult Drosophila optic lobe scRNA-seq dataset²⁴. Neuronal clusters are annotated with cell type identities and positive marker combinations. j Seven neuron types are unexpectedly born within the Vsx1–Hth spatial–temporal window. In all images and schematics: Anterior is left. Dorsal is up in (a). Scale bar: 15 µm.

To identify the cell types labeled by the vsx1⋂hth-Gal4 driver, we used the MultiColor FlpOut (MCFO)³⁹ system to generate single-cell clones of neurons in the adult brain. Remarkably, we found that six morphologically distinct neuronal cell types are labeled by the split-Gal4: Pm3, TmY15, TmY17, Tm23, Li2, and TmY12 (Fig. 1c–h and Supplementary Fig. 1d–i). The morphology of five of these neurons (Pm3, TmY15, Tm23, Li2, and TmY12) has previously been described^14,40,41,42. TmY17 neurons, which have not been reported, are similar to TmY15 in morphology but do not arborize in the distal medulla and send a projection to only the most superficial lobula plate layer (Fig. 1d, e and Supplementary Fig. 1e, f)^40,41.

We next determined whether the six neuronal cell types labeled by vsx1⋂hth-Gal4 are discoverable in a single-cell RNAseq dataset of the adult optic lobe (see “Data availability”)²⁴. Cells in clusters marked by upregulated levels of vsx1 and hth were subsetted and independently re-clustered, leading to the emergence of five distinct clusters (Supplementary Fig. 2a). Using cluster-specific marker combinations, we identified sets of markers that differentiate the neuron types labeled by the vsx1⋂hth-Gal4 driver line and were therefore able to assign each of the five clusters a cell type identity (Fig. 1i and Supplementary Fig. 2b–t). Through this analysis, we also found that vsx1⋂hth-Gal4 labels two morphologically indistinguishable sub-types of Pm3 neurons, henceforth referred to as Pm3a and Pm3b (Supplementary Fig. 2i–l). The only neurons labeled by the vsx1⋂hth-Gal4 line that we were unable to identify in the single-cell RNAseq dataset were Li2 and TmY12. We posit that these two neuronal cell types are present in too low a number in the adult optic lobe to form identifiable clusters in the single-cell RNAseq dataset.

Taken together, the above clonal and single-cell RNAseq analyses demonstrate that not one but seven neuronal types are generated at the intersection of the Vsx1 spatial and Hth temporal axes. These neurons can be labeled by the following marker combinations: Pm3a (AstA⁺/E93⁺/Zfh1⁺), Pm3b (Svp⁺/Zfh1⁺), Tm23 (Mamo⁺⁺), TmY17 (Zfh1⁺⁺/Mamo⁺/Br⁺), TmY15 (Br⁺), Li2 and TmY12 (no markers identified) (Fig. 1i, j and Supplementary Fig. 2t). The unexpected neuronal diversity generated at this spatial–temporal address suggests that an unidentified mechanism exists (in addition to spatial and temporal patterning) to further diversify cell fates in the medulla.

Vsx1–Hth neuronal cell types are sequentially born over the three days of medulla neurogenesis

In our analysis of Vsx1–Hth neurons, we observed that cell bodies of the same neuronal type are localized to restricted regions along the A–P axis of the medulla cortex: Li2 cell bodies occupy the anterior region of the cortex, followed by (from anterior to posterior) the cell bodies of TmY17, TmY15, Tm23, Pm3b, Pm3a and TmY12 neurons (Fig. 2a and b). In contrast to their restricted A–P position, Vsx1–Hth neuronal cell types are uniformly dispersed along the D–V axis (Fig. 2b), an observation that is consistent with the previous finding that Vsx1 neurons migrate dorsally and ventrally from their central OPC birth location during pupal development²³.

Fig. 2: Vsx1–Hth neuronal cell bodies occupy distinct regions along the A–P axis of the adult medulla cortex, correlating to their sequential birth order in the larva.

a and b Schematics depicting cell body positions of Vsx1–Hth neurons in the medulla cortex as visualized in horizontal (a) and frontal (b) orientations of the optic lobe. c–e Early-born medulla neurons generated from 48 h ALH to the cessation of larval feeding (92 h ALH) are labeled by EdU (red, d). Late-born neurons generated after the feeding window (96 h ALH to 15 h APF) are labeled by memory trace experiments (βGal, green, e). Arrows indicate the extent of neurons labeled across the A–P axis by EdU feeding (d) and memory trace (e) experiments. f–i 48–92 h ALH EdU feeds (red) label Zfh1⁺⁺ TmY17 neurons (cyan, f) and Br⁺ TmY15 neurons (blue, g), but not Mamo⁺⁺ Tm23 neurons (white, h) or Zfh1⁺ Pm3 neurons (cyan, i). j and k 72–78 h ALH EdU feeds (red) label Zfh1⁺⁺ TmY17 neurons (magenta, j), but not Zfh1^- TmY15 neurons (k). l and m 84–90 h ALH EdU feeds (red) do not label Zfh1⁺⁺ TmY17 neurons (magenta, l), but do label Zfh1^- TmY15 neurons (m). n Quantification of (j–m). The percentage of Vsx1–Hth EdU⁺ cells in the adult identified as TmY17, TmY15, Tm23, and Pm3 was quantified for 6 h EdU feeds at four developmental stages (63–69 h ALH n = 8 neurons; 72–78 h ALH n = 37 neurons; 78–84 h ALH n = 23 neurons; 84–90 h ALH n = 35 neurons). Error bars denote SEP. Source data are provided as a Source Data file. o Quantification of Supplementary Fig. 4a–j. The percent of TmY17, TmY15, Tm23, and Pm3 neurons labeled by a pxb-Gal4 memory trace after heat shock induction at four different developmental stages (86 h ALH n = 439 Pm3s, 29 Tm23s and 653 TmY15s; 92 h ALH n = 664 Pm3s, 31 Tm23s and 578 TmY15s; 0 h APF n = 586 Pm3s and 46 Tm23s; 10 h APF n = 476 Pm3s). Error bars denote SEP. Source data are provided as a Source Data file. p Schematic displaying the birth windows of Vsx1–Hth neuron types over the course of neurogenesis based on EdU and genetic birthdating analysis. q Model illustrating the relationship between neuronal birth order and proneural wave propagation in the medial (anterior) to lateral (posterior) direction during medulla neurogenesis. In all images and schematics: Anterior/medial is left. Dorsal is up in (b). White arrowheads indicate neuronal cell bodies in (f–m). Scale bar: 15 µm.

We asked how the striking A–P distribution of Vsx1–Hth cell types in the adult medulla cortex is established during development. Over the span of three days, initiating at 48 hours (h) after larval hatching (ALH) and ending at 15 h after puparium formation (APF), a proneural wave converts OPC NE cells into NBs in a medial to lateral direction^15,19,20,28. A consequence of this moving wave is that neurons born early in development occupy a medial position in the developing cortex (anterior in the adult), whereas later-born neurons are located in the lateral region (posterior in the adult)⁴³. To determine whether the A–P position of cell bodies in the adult cortex corresponds to their time of birth during development, we used EdU labeling and heat shock-induced lineage tracing to birthdate medulla neurons (Fig. 2c). Rearing larvae on EdU-treated food labels all neurons born between the start of neurogenesis and the cessation of larval feeding at ~92 h ALH. Remarkably, we found that the majority of EdU⁺ cell bodies are restricted to the anterior half of the medulla cortex (Fig. 2d), which suggests that there is limited cell body movement along the A–P axis during pupal development. In shorter feeding intervals, the timing of EdU administration correlates to the A–P position of labeled cells in the adult cortex; earlier feeds label more anterior cells, whereas later feeds label more posterior cells (Supplementary Fig. 3b). To birthdate neurons born after 92 h ALH, we utilized a lineage-based approach in which neurons born in the Vsx1 compartment are permanently labeled by βgal after heat shock induction (Fig. 2c and Supplementary Fig. 3a). We found that a heat shock at 92 h ALH resulted in the labeling of cell bodies in the posterior half of the adult cortex, with only the occasional βgal⁺ cell located anteriorly (Fig. 2e). Heat shocks at later time points progressively labeled more posterior cells in the cortex (Supplementary Fig. 3c). The limited movement of cell bodies observed along the A–P axis in our birthdating analyses indicates that the A–P position of neurons in the adult medulla cortex is correlated to their time of birth during neurogenesis.

We next birthdated the neurons labeled by vsx1⋂hth-Gal4 > GFP and found that, consistent with their anterior cell-body positions, only TmY17 and TmY15 neurons incorporate EdU after larval feeding; the more anterior TmY17 neurons are labeled by early feeds and the posteriorly adjacent TmY15 neurons are labeled by late feeds (Fig. 2f–n). Conversely, the posterior Tm23 and Pm3 neurons are only labeled by the lineage trace birthdating technique; heat shock induction at 92 h ALH labels both neuron types, whereas only the more posterior Pm3 neurons are labeled after heat shock induction in the pupa, indicating that Pm3 neurons are generated after Tm23 neurons (Fig. 2o and Supplementary Fig. 4a–j). Taken together, our birthdating analyses demonstrate that Vsx1–Hth neuronal types are sequentially born during medulla neurogenesis: TmY17 neurons are generated first, followed by TmY15, Tm23, and Pm3 neurons (Fig. 2p). The sequential birth order of these neurons is thus correlated with their cell body positions along the A–P axis of the adult medulla, falling in line with the progression of the proneural wave in the medial-to-lateral direction during neurogenesis (Fig. 2a, b and q). While our birthdating analysis did not discriminate between the two Pm3 subtypes, the A–P distribution of these neurons suggests that Pm3b neurons are generated before Pm3a (Fig. 2a, b and q). Similarly, the A–P cell body positions of Li2 and TmY12 neurons suggest that Li2 neurons are born earliest, whereas TmY12 neurons are generated last (Fig. 2a, b and q). These findings indicate that Vsx1–Hth NBs make distinct neuronal types based on when during the three days of neurogenesis they undergo the NE–NB transition. We next asked whether the temporal patterning of the OPC NE from which the NBs are derived could account for this neuronal diversity.

The OPC NE is temporally patterned by opposing gradients of the Imp and Syp RNA-binding proteins

A search for patterning factors that are differentially expressed in the OPC NE over the course of medulla neurogenesis identified the Imp and Syp RNA-binding proteins. Imp and Syp have been previously shown to function as temporal patterning factors in the NB lineages of the central brain, where the two proteins act to diversify neuronal fates via the post-transcriptional regulation of downstream TFs^{35,36,37,44,45}. We found that opposing temporal gradients of Imp and Syp pattern the OPC NE over the course of neurogenesis; Imp is expressed at high levels in the NE at 48 h ALH (early third instar) and declines in a gradient manner over time, becoming undetectable at 96 h ALH (late third instar) (Fig. 3a–d and i), whereas Syp expression is absent at 48 h ALH and increases over time, resulting in high levels of expression at the end of neurogenesis (Fig. 3a–d and j). Imp and Syp are also expressed in the medulla NBs that are generated from the NE, with Imp levels high in early third instar NBs and Syp levels high in late third instar NBs (Fig. 3e–h).

Fig. 3: Opposing gradients of Imp and Syp temporally pattern the OPC NE.

a–c Imp (green) and Syp (red) expression in the early (a), mid (b) and late third instar (c) larval brain. White dashed lines outline the OPC NE, yellow arrowheads indicate optic lobe (OL) NBs, and white arrowheads indicate central brain (CB) NBs. Schematics (a”’, b”’, c”’) have been created in BioRender. d Quantification of Imp (green) and Syp (red) intensities in the OPC NE from 48 to 96 h ALH. Plot illustrates the mean relative fluorescence intensity of Imp and Syp at each time point (48 h ALH n = 5 optic lobes; 60 h ALH n = 7 optic lobes; 72 h ALH n = 5 optic lobes; 84 h ALH n = 12 optic lobes; 96 h ALH n = 10 optic lobes). Error bars denote SEM. Source data are provided as a Source Data file. e–h Expression of Imp (green) and Syp (red) in the OPC NE (Arm, gray), and NBs (Dpn, magenta) in the early (e and f) and late third instar (g and h) larva. Yellow arrows denote the propagation direction of the proneural wave, and white arrowheads indicate NBs. i–n Imp (green, i), Syp (red, j), Chinmo (cyan, k), Mamo (yellow, l), Br (blue, m) and E93 (magenta, n) expression in the OPC NE (white dashed lines) at different developmental stages. Optic lobe neurons (asterisk) and the youngest neuroblast (arrowheads) are also labeled. o Schematic summary of the Imp, Syp and NE tTF expression windows in the OPC NE during medulla neurogenesis. p Schematic modeling how temporal gradients of Imp (green) and Syp (red) assign unique identities to NBs and neurons over the course of neurogenesis. Neurons are pseudo-colored based on the stage of neurogenesis in which they were generated. In all images and schematics: Medial is left. Dorsal is up in (a–c). Scale bar: 15 µm.

We next analyzed the expression of candidate TFs that are regulated by Imp and Syp in central brain NB lineages^36,37,44 and found four TFs that are temporally expressed in the OPC NE: Chinmo, Mamo, Broad and E93. Over the course of neurogenesis, expression of these TFs defines at least four temporal windows in the OPC NE: a Chinmo⁺⁺/Mamo⁻/Br⁻/E93⁻ window from 48 to 60 h ALH, a Chinmo⁺/Mamo⁺⁺/Br⁺/E93⁻ window from 60 to 84 h ALH, a Chinmo⁻/Mamo⁺/Br⁺/E93⁻ window from 84 to 96 h ALH, and a Chinmo⁻/Mamo⁻/Br⁺/E93⁺ window from 96 h ALH to 15 h APF (Fig. 3i–o and Supplementary Fig. 5a–i). The expression of Chinmo and Mamo in declining gradients may act to further subdivide the early windows. To differentiate these temporally expressed NE TFs from the previously identified tTFs that pattern the NBs as they age, we will henceforth refer to Chinmo, Mamo, Br and E93 as NE tTFs, and the NB temporal patterning factors as NB tTFs.

We found that the NE tTFs are expressed in OPC NBs and neurons as well. In NBs, the NE tTFs are expressed in temporal windows that align with their NE expression, with the exception of Mamo, which continues to be expressed in NBs for an additional 12 h after its NE expression has declined (Fig. 3k–n). In the neurons of the medulla cortex, we observed that the NE tTFs are expressed in subsets of neurons along the medial–lateral axis that are consistent with their temporal order of expression in the NE. Chinmo is expressed in the earliest born, most medial neurons, followed by Mamo and then E93 in the latest born, most lateral neurons (Supplementary Fig. 5j–l). We observed a similar Chinmo-Mamo-Broad-E93 order of expression along the medial–lateral axis in Vsx1–Hth neurons as well (Supplementary Fig. 5m–o). The temporal expression of Imp, Syp and the NE tTFs in the OPC makes these factors excellent candidates to diversify neuronal fates in the medulla. As the proneural wave sweeps across the NE, newly formed NBs and their neuronal progeny may be assigned unique identities based on the levels of these temporal factors in the NE (Fig. 3p).

Temporal patterning of the OPC NE diversifies Vsx1–Hth neuronal fates

We next asked whether the temporal patterning of the OPC NE contributes to the generation of neuronal diversity in the medulla. As a first step, we determined whether Imp and Syp are required for the expression of the NE tTFs. In an imp loss-of-function (LOF) mutant, in which Imp expression is lost specifically in the OPC NE (Supplementary Fig. 6a and b), we found that expression of the early NE tTFs Chinmo and Mamo is unaffected (Fig. 4a–d). However, the late NE tTF E93 is ectopically expressed, appearing in the NE ~ 24 h earlier than it does in the wild-type NE (Fig. 4g and h). In syp^RNAi LOF clones (Supplementary Fig. 6c), we found that expression of the early NE tTFs Chinmo and Mamo is derepressed into the mid and late third instar NE, respectively, while expression of the late factor E93 is lost (Fig. 4j–l). Additionally, we observed that, unlike what has been shown in central brain NBs^35,36, Imp and Syp do not negatively cross-regulate each other in the OPC NE (Fig. 4e, f and i). Taken together, these LOF data demonstrate that Imp is dispensable for the expression of the early NE tTFs but is required to repress the late NE tTF, E93. In contrast, Syp acts to repress the expression of the early NE tTFs and promote the late expression of E93 (Fig. 4m).

Fig. 4: NE temporal patterning diversifies Vsx1–Hth neuron fates.

a–h Chinmo (green, a and b), Mamo (green, c and d), Syp (green, e and f) and E93 (green, g and h) expression in the OPC NE (Arm or DE-Cad, gray) in wild-type and imp LOF brains at the mid-third instar stage. i–l Imp (green, i), Chinmo (green, j), Mamo (green, k), and E93 (green, l) in syp^RNAi LOF clones (RFP, red) made in the OPC NE at different developmental stages. White dashed lines outline the boundary of the clone. m Schematic summarizing the regulatory relationships between Imp, Syp and the NE tTFs in the OPC NE during neurogenesis. n–s Max projection images of vsx1∩hth-Gal4 > GFP (green) neurons and cell-type-specific markers in the wild-type (n and o), imp LOF (p and q), and syp^RNAi LOF (r and s) adult medulla. Zfh1⁺ TmY17 neurons (red, n, p and r) are outlined with white dashed circles. A subset of AstA⁺ Pm3a neurons (magenta, o, q and s) is indicated with arrows. t Schematic summary of the data presented in (n–s). u–x Max projection images of vsx1∩hth-Gal4 > GFP (green) neurons and cell-type-specific markers in the wild-type (u and v) and e93^RNAi LOF (w and x) adult medulla. Mamo⁺⁺ Tm23 neurons (red, u and w) are indicated with white arrowheads. A subset of AstA⁺ Pm3a neurons (magenta, v and x) is indicated with white arrowheads. y Schematic summary of the data presented in (u–x). z Quantification of the mean number of TmY17, TmY15, Tm23 and Pm3a neurons in wild-type and LOF mutant backgrounds. Asterisks illustrate a significant difference in neuron number between the wild-type and LOF optic lobes (two-sided t-test, P < 0.05; wild-type n = 40 optic lobes; imp LOF n = 15 optic lobes; syp LOF n = 34 optic lobes; mamo LOF n = 21 optic lobes; e93 LOF n = 15 optic lobes). Error bars denote SEM. Source data are provided as a Source Data file. In (n–y): Dorsal is up and anterior is left. Scale bar: 15 µm.

We next asked whether Imp, Syp and the NE tTFs are required for the specification of Vsx1–Hth neuronal fates. We used the imp LOF mutant and vsx1⋂hth-Gal4 > RNAi to knock-down the expression of the NE temporal patterning genes and analyzed cell type-specific markers in adult brains. In the imp LOF mutant medulla, early-born TmY17 neurons are lost, intermediate-born TmY15 and Tm23 neurons are significantly reduced in number, and late-born Pm3a neurons are unaffected (Fig. 4n–q, t and z). Conversely, in syp knock-down mutants, TmY17 neurons nearly double in number, whereas the number of Pm3a neurons is significantly reduced (Fig. 4n–o, r–t and z). Imp and Syp temporal patterning is thus required to diversify the fates of Vsx1–Hth neurons, with Imp promoting early fates and Syp promoting late ones. We next analyzed the NE tTFs and found that knock-down of the late gene e93 led to the loss of late-born Pm3a neurons and an increase in the number of earlier-born TmY17 and Tm23 neurons (Fig. 4u–z and Supplementary Fig. 6f). Knock-down of the early gene mamo had a more limited effect, only resulting in a decrease in TmY15 neurons (Fig. 4z and Supplementary Fig. 6d, e). Indeed, we found that TmY15 neuronal number was decreased in all LOF backgrounds (Fig. 4z), which suggests that this cell type may require the function of both the early and late temporal patterning factors for its specification. Finally, we were unable to analyze the role of Chinmo in fate specification as vsx1⋂hth-Gal4>chinmo^RNAi resulted in larval lethality.

The above data demonstrate that Imp, Syp and the downstream NE tTFs are required for the diversification of Vsx1–Hth neuronal fates. The temporal patterning of the OPC NE thus represents a third major patterning mechanism, which acts together with the previously identified NE spatial and NB temporal inputs to generate neuronal diversity in the medulla.

NE temporal patterning extends to multiple spatial and NB temporal windows

Does Imp and Syp patterning extend beyond Vsx1 and Hth to include other spatial–temporal birth windows? We first analyzed the expression of Imp and Syp in OPC NBs and found that Imp is expressed in all NBs at the early third instar stage, while Syp is expressed in all NBs at later stages (Fig. 5a–d), which suggests that Imp and Syp patterning extends beyond the Hth NB tTF window. We next asked whether the NB tTF series is present in OPC NBs across multiple stages of neurogenesis and observed that the early NB tTF Extradenticle (Exd), the intermediate NB tTF Ey, and the late NB tTF Slp are sequentially expressed in medulla NBs at 72 h ALH, 96 h ALH, 0 h APF and 5 h APF (Fig. 5e–h). The expression of Imp and Syp in all NBs at early and late stages of neurogenesis, respectively, together with the presence of the NB tTF series in multiple stages of neurogenesis, suggests that these two patterning mechanisms work concurrently to diversify neuronal fates. As Imp and Syp are expressed in the NE from which the NBs are derived, we next asked if the NB tTF series requires Imp and Syp for its progression. We found that the NB tTF series progresses independently of Imp and Syp temporal patterning; the NB tTFs, Exd, Ey and Slp, are expressed normally in late third instar larval brains when either imp or syp expression is lost (Fig. 5i–l). Taken together, these data support a model in which both the NE and NB temporal patterning mechanisms work concurrently and independently to diversify neuronal fates throughout the medulla.

Fig. 5: NE temporal patterning extends to other OPC spatial compartments and temporal windows.

a–d Imp (green, a and c) and Syp (red, b and d) expression in OPC NBs (Dpn, gray) of the early and late third instar larval brain. e–h Exd (red), Ey (green), and Slp2 (blue) in OPC NBs at 72 h ALH (e), 96 h ALH (f), 0 h APF (g), and 5 h APF (h). Gray dashed line marks the boundary between the OPC NE and NBs. White arrow illustrates the progression of the NB tTF series. i and j Expression of Exd (red, i), Ey (green, j) and Slp2 (blue, j) in OPC NBs (Dpn, gray) of an imp LOF mutant. k and l Expression of Exd (red, k), Ey (green, l) and Slp2 (blue, l) in syp^RNAi LOF NBs. syp^RNAi is driven in cOPC NBs by MzVUM-Gal4. The cOPC (GFP, gray) boundary is marked by white dashed lines. m–v The developmental birth origins of Dm4 (m–q) and Tm26 (r–v) neurons. Schematics illustrating the spatial–temporal birth origin (m and r), morphology (o and t), and cell body positions (p and u) of Dm4 and Tm26 neurons are on the left. Dm4 and Tm26 neurons are labeled in the adult optic lobe (Brp, magenta) by Dm4-Gal4 > GFP (green, n) and Tm26-Gal4 > GFP (green, s), respectively. White arrowheads denote neuronal cell bodies and white dashed lines mark medulla layer M7. 60–92 h ALH EdU feeds (red, q and v) label Dm4 (white dashed circles, q”) but not Tm26 (white dashed circles, v”) neurons. Region of the medulla cortex outlined with white dashed lines (q and v) is magnified in the images below. w The proportion of Dm4 neurons (green) and Tm26 neurons (red) labeled by EdU administered from 60–92 h ALH (two-sided two-proportion Z-test, P < 0.00001, n = 409 Dm4 neurons and 164 Tm26 neurons). Error bars denote SEP. Source data are provided as a Source Data file. x Schematic depicting the developmental birth windows of Dm4 and Tm26 neurons. In all images and schematics: Anterior/medial is left. Dorsal is up in (a–l, m, p, r, u and x). Scale bar: 15 µm.

We also observed that Imp, Syp and the NE tTFs are expressed in all spatial compartments of the OPC NE, which suggests that NE temporal patterning may extend to all spatial–temporal neuronal birth addresses (Supplementary Fig. 7a and b). To determine whether neuron types generated outside of the Vsx1 spatial and Hth temporal axes are also born in restricted NE temporal windows, we mapped the birthdate of two additional multi-columnar neurons, Dm4 and Tm26. Using a combination of lineage trace and marker analyses, we found that both Dm4 and Tm26 neurons are derived from the ventral-Optix spatial compartment and Ey temporal window (Fig. 5m–o, r–t and Supplementary Fig. 7c–h)^31,33. In the adult medulla cortex, we found that Dm4 cell bodies are located in the anterior region while Tm26 cell bodies are restricted to the posterior, suggesting that these neurons are generated in early and late developmental windows, respectively (Fig. 5p and u). EdU birthdating analysis confirmed that Dm4 neurons are generated early in development; EdU feeds between 60 and 92 h ALH labeled over 90% of Dm4 neurons (Fig. 5q and w). In contrast, 60–92 h ALH EdU feeds failed to label Tm26 neurons, confirming that these neurons are generated in a late developmental window (Fig. 5v and w). The mapping of neurons from the same spatial–temporal address to distinct NE temporal windows, Dm4 (early) and Tm26 (late), suggests that Imp and Syp patterning extends beyond Vsx1 and Hth to include the Optix spatial and Ey temporal windows (Fig. 5x). NE temporal patterning thus likely acts as a global patterning mechanism to generate diversity in the medulla.

Uni-columnar neurons are specified independently of Imp and Syp patterning

We next asked whether NE temporal patterning extends to the specification of uni-columnar neurons in the medulla. It has previously been shown that uni-columnar neurons, which contact only one medulla column and are thus generated in larger numbers than multi-columnar neurons, are specified throughout the D–V axis of the OPC, independently of spatial patterning²³. We analyzed the cell body positions of two uni-columnar neurons, Mi1 and Tm3, and found that they are distributed evenly throughout both the A–P and D–V axes of the adult medulla cortex (Fig. 6a–d). The even distribution of these neurons across the A–P axis suggests that uni-columnar neurons are generated throughout the 3 days of neurogenesis. Indeed, EdU birthdating demonstrates that Mi1 and Tm3 neurons are generated in both early (EdU feeding) and late (post-EdU feeding) developmental windows (Fig. 6e and f). Additionally, we found that the Mi1-specific marker, Bsh, is expressed in newly generated Mi1 neurons in the early, intermediate, and late stages of neurogenesis, indicating that Mi1 neurons are likely born in all NE temporal windows (Fig. 6g–j).

Fig. 6: Uni-columnar neurons are specified independently of NE temporal patterning.

a Schematic illustrating the spatial–temporal birth origin of Mi1 (green) and Tm3 (orange) uni-columnar neurons in the larval OPC. b Schematic depictions of Mi1 and Tm3 neurons in the optic lobe. c and d Schematics displaying Mi1 (c) and Tm3 (d) cell body positions in the medulla cortex. e 60–92 h ALH EdU feeds (red) label Mi1 neurons (Bsh, green) in the anterior (e’–e”), but not posterior (e”’–e””) medulla cortex. Cell bodies are outlined with white dashed circles. f 70–78 h ALH EdU feeds (red) label Tm3 neurons (Tm3-Gal4 > GFP, green) in the anterior (f’–f”), but not posterior (f”’–f””) medulla cortex. Cell bodies are outlined with white dashed circles. g–j Mi1 neurons (Bsh, green) in the developing optic lobe at different stages of neurogenesis. Arm (magenta) labels the OPC NE, and white arrows indicate the direction of proneural wave propagation. The youngest Mi1 neurons at each stage are denoted with a yellow arrowhead. Lamina neurons are also labeled by Bsh (white arrow, i and j). k and l Mi1 neurons (Bsh, green) in a wild-type (k) and imp LOF (l) optic lobe at 5 h APF. DE-Cad (gray) labels the OPC NE. m Mi1 neurons (Bsh, green) in syp^RNAi LOF clones (RFP, red) at 5 h APF. Yellow dashed lines mark the boundary of the clone. n Schematic depicting the developmental birth windows of Mi1 and Tm3 neurons. In all images: Medial/anterior is left. Dorsal is up in (a, c, d and n). Scale bar: 15 µm.

The observation that Mi1 and Tm3 neurons are generated in all NE temporal windows suggests that uni-columnar neurons do not require Imp and Syp patterning for their specification (Fig. 6n). Indeed, we found that the medial–lateral distribution of Mi1 neurons in the early pupal cortex is unaffected in both the imp LOF OPC and syp^RNAi LOF clones (Fig. 6k–m). Taken together with the previous observation that uni-columnar neurons are generated independently of spatial patterning inputs²³, the above data demonstrate that uni-columnar neurons rely solely on NB tTF patterning for their specification.

NE temporal patterning generates unexpected A–P regionalization of the medulla neuropil

The medulla neuropil is made up of ~800 columns that receive retinotopic inputs from the photoreceptors in the overlying ommatidia^18,46. The restricted cell body positions of multi-columnar neurons along the A–P axis led us to investigate whether the arborizations of these neurons innervate all columns in the neuropil. We first analyzed Pm3a innervations using the AstA antibody, which labels Pm3a cell bodies and neurites⁴⁷, and found that despite their posteriorly restricted cell bodies, Pm3a arborizations extend across the entire A–P axis of the neuropil (Fig. 7a). Similar results were observed for the anteriorly restricted TmY15 and Dm4 neurons, whose projections in the distal medulla also cover the A–P axis of the neuropil (Fig. 7b and c). Our findings for Pm3a, TmY15 and Dm4 are not surprising, as multi-columnar medulla neurons have previously been shown to innervate all medulla columns, with the exception of a select few neuron types such as Dm-DRA1 and Dm-DRA2 (_DRADm8), which only innervate the dorsal-most columns^{14,48,49,50,51}. Surprisingly, however, we found that three multi-columnar cell types — Tm26, TmY12 and TmY17 — do not innervate all medulla columns, but rather only those closest to their cell body locations. Tm26 arborizations are restricted to the posterior–ventral third of columns, TmY12 neurons innervate the most posterior row of columns, and TmY17 neurons send arborizations to columns in the anterior–ventral region of the neuropil (Fig. 7d–g and i–k). The regionalization of Tm26, TmY17 and TmY12 innervations indicates that the medulla neuropil contains unanticipated patterning of its retinotopic map along the A–P axis (Fig. 7m and n).

Fig. 7: NE temporal patterning regionalizes the medulla neuropil across the A–P axis.

a–d Pm3a (AstA, green, a), TmY15 (vsx1∩hth-Gal4 > GFP, green, b), Dm4 (Dm4-Gal4 > GFP, green, c) and Tm26 (Tm26-Gal4 > GFP, green, d) neuronal projections in the medulla neuropil (Brp, red). Arrowheads denote the medulla neuropil layer(s) innervated by each neuron type. e–h Schematics depicting the cell body positions of Tm26 (e), TmY12 (f), TmY17 (g), and TmY17 in a vsx1∩hth-Gal4>syp^RNAi LOF background (h). 3D reconstructions of representative MCFO-labeled clones (V5, green) of each neuron type in the adult medulla (Chp, magenta) are on the right (e’, f’, g’, h’). i–l The distribution of medulla neuropil innervations observed in MCFO-labeled clones of Tm26 (i, n = 24 clones), TmY12 (j, n = 7 clones), TmY17 (k, n = 15 clones), and TmY17 in a vsx1∩hth-Gal4>syp^RNAi LOF background (l, n = 19 clones). Projection traces of individual neuronal clones in the medulla neuropil are on the left. Heatmaps representing the proportion of total clones innervating a given spatial region of the medulla along the A–P and D–V axis (color intensity, white-to-red) are on the right. Source data are provided as a Source Data file. m Schematic illustrating the propagation of visual information from the overlying eye and through the optic lobe neuropils (gray arrows). TmY17 (blue), TmY12 (magenta) and Tm26 (yellow) neurons process information received exclusively from the anterior or posterior regions of the visual field. n Schematic illustrating the region of the compound eye from which TmY17 (blue), TmY12 (magenta) and Tm26 (yellow) neurons receive visual information. In all images and schematics: Anterior is left. Dorsal is up in (e–l and n). Scale bar: 15 µm.

We next determined whether NE temporal patterning is required for the regionalized innervation pattern of TmY17 neurons. As noted above, the number of TmY17 neurons is significantly increased in the syp^RNAi LOF background (Fig. 4n, r and t). Consistent with the window of TmY17 production lasting longer in the absence of Syp expression, the cell body positions for these neurons occupy a larger region along the A–P axis, extending into the posterior half of the cortex (Fig. 7h). This observation led us to determine whether the innervations of TmY17 neurons in the syp^RNAi LOF background also expand into the more posterior neuropil columns, or alternatively, remain restricted to anterior columns like their wild-type counterparts. To distinguish between these two possibilities, we generated vsx1⋂hth-Gal4>syp^RNAi MCFO clones and mapped the projections of TmY17 neurons in the neuropil. We observed that the arborizations of syp^RNAi LOF TmY17 neurons cover a significantly larger A–P region of the medulla neuropil compared to wild-type, extending innervations to posterior columns (Fig. 7k and l). The extent of the posterior innervations matches the posterior expansion in cell body position observed for syp^RNAi LOF TmY17 neurons, which suggests that Syp contributes to the regionalization of the medulla neuropil via the positioning of cell bodies along the A–P axis of the cortex.

Discussion

In this study, we identify temporal patterning of the NE as a third major patterning mechanism in the OPC that acts together with the previously reported spatial and NB temporal axes to diversify medulla neuronal fates (Fig. 8a and b). We find that opposing gradients of the Imp and Syp RNA-binding proteins, together with their downstream NE tTFs, temporally pattern the OPC NE over the course of medulla neurogenesis. As proof of principle, we demonstrate that NBs generated at the intersection of one spatial–temporal address, Vsx1–Hth, generate not one but seven distinct neuronal cell types and that NE temporal patterning is required for their specification. We further show that Imp and Syp are expressed in all spatial compartments and NB temporal windows, and that neurons born at a second spatial–temporal address, Optix—Ey, are also generated in distinct NE temporal windows, which suggests that Imp and Syp patterning may diversify neuronal fates across all spatial–temporal addresses in the OPC.

Fig. 8: Inputs across three patterning axes in the OPC are integrated to generate neural diversity.

a The developmental birth windows of uni- and multi-columnar neurons over the course of neurogenesis. b A combination of NE temporal (i.e., Imp, Syp), NB temporal (i.e., Hth, Ey, Slp, D, Tll), spatial (not pictured) and binary Notch status contributes to the generation of neural diversity in the medulla. c Concurrent temporal patterning mechanisms can act to generate neural diversity in systems with distinct symmetric (produce two progenitors; P–P) and asymmetric (produce a progenitor and neuron; P–N) progenitor division modes.

The identification of a third patterning axis significantly increases the potential amount of neuronal diversity that OPC NBs can generate in the medulla. The combination of 12 spatial compartments, 11 NB temporal windows and the Notch-based binary diversification of neuronal fates can generate up to 264 distinct cell types^{21,23,31,32,33,34,42}. Based on our observation that NE temporal patterning results in a 7-fold increase in neural diversity at a single spatial–temporal address, we propose that inputs across the three patterning axes, combined with Notch-based diversification, can produce up to 1848 medulla cell types. This number is likely a significant overestimate of the actual diversity present in the medulla, as the programmed cell death of Notch^ON or Notch^OFF GMC outputs, as observed for tOPC-derived medulla neurons and ventral nerve cord neurons, will likely greatly reduce the number of cell types generated^29,52,53. Furthermore, our finding that uni-columnar medulla neurons are generated independently of spatial²³ and NE temporal inputs will also decrease the number of neuron types produced. However, the identification of a third patterning axis suggests that the discovery of ~100 neuronal cell types in the adult medulla via single-cell RNAseq²⁴ has not captured the full breadth of diversity within this system. Indeed, recent electron microscopy-based reconstructions of the Drosophila brain have identified ~50 additional medulla cell types that had not been previously reported^25,49,50. Future studies should extend the analysis of Imp-Syp patterning to additional spatial–temporal addresses to determine the full extent of diversity generated by this newly identified patterning mechanism in the OPC.

Future studies should also determine whether additional NE tTFs act downstream of Imp and Syp in the NE. While we show that the expression of the NE tTFs, Chinmo, Mamo, Broad and E93, accounts for four distinct temporal windows, our identification of seven Vsx1–Hth neuron types suggests that additional factors play a role in the diversification of fates by this patterning axis. Moreover, our finding that Imp regulates early fates independent of Chinmo and Mamo suggests that an additional NE tTF functions in the early NE. The use of single-cell RNAseq and proteomic approaches at multiple time points during medulla neurogenesis should identify additional NE temporal patterning genes. Furthermore, the use of single-cell ATACseq will provide insights into how all three patterning mechanisms (spatial, NE temporal, NB temporal) are integrated in OPC progenitors at both the transcriptional and epigenetic levels to specify neural fates. It is anticipated that the genes and mechanisms identified downstream of Imp and Syp in the OPC will also play a role in the central brain and ventral nerve cord, where Imp and Syp temporally pattern both Type I and Type II NB lineages^35,36,37,45.

Our findings indicate that medulla NBs are temporally patterned by both the Imp/Syp state of the NE from which each NB is generated and the tTF cascade that progresses within the NB as it ages. Combinatorial temporal patterning of neural progenitors has been previously reported in the Type II NBs of the Drosophila central brain, in which NBs are patterned by Imp and Syp, while their daughter cells, the intermediate neural progenitors, are patterned by a tTF cascade^45,54. One notable difference between Imp/Syp patterning in the central brain Type II NBs and the OPC NE is the timing of the Imp-to-Syp transition. In the OPC NE, the Imp-to-Syp transition occurs at 72 h ALH, whereas in central brain Type II NBs, this transition occurs earlier at 60 h ALH^36,55. While work in the central brain has shown that temporal mechanisms can pattern two types of asymmetrically dividing progenitors in a single lineage (NB and intermediate neural progenitor), here we demonstrate that concurrent temporal mechanisms can also pattern the symmetric (NE) versus asymmetric (NB) stages of a progenitor lineage to diversify neuronal fates. We anticipate that this mode of concurrent temporal patterning will be especially relevant to the study of neural stem cell lineages in the vertebrate brain, where the majority of progenitors undergo a similar transition from symmetric divisions that expand the progenitor pool to asymmetric divisions that generate neurons⁵⁶. The temporal patterning of vertebrate progenitors in both the symmetric and asymmetric stages of their lineage would greatly increase the amount of neuronal diversity that can be generated and would thus help to account for the extensive diversity found in complex brain structures (Fig. 8c). For example, in the developing vertebrate retina, retinal progenitor cells (RPCs) transition from symmetric to asymmetric divisions during retinogenesis⁵⁷. While the temporal patterning of RPCs has been described as a way to generate the birth order-dependent sequence of neuronal classes in the retina^58,59,60,61, current models have not been able to fully account for the diversity of cell types found within each class⁵⁷. Our findings raise the possibility that the independent temporal patterning of RPCs in their symmetric versus asymmetric modes of division could account for both the sequential production of neuronal classes observed over time (i.e. early amacrine vs late bipolar cell classes), as well as the subtype diversity observed within each class (i.e. 15 types of bipolar cells). Indeed, it has recently been shown that late RPCs generate distinct bipolar cell subtypes in a defined temporal order, which suggests that a second intrinsic temporal patterning mechanism may act in asymmetrically dividing RPCs to regulate the sequential generation of bipolar cell subtypes⁶².

In this study, we also demonstrate that the temporal patterning of the NE, combined with the uni-directional progression of neurogenesis, leads to the birthdate-dependent distribution of neurons along the A–P axis of the adult medulla cortex. This finding indicates that, unlike the extensive migration of neurons observed along the D–V axis during pupal development²³, there is limited migration of neurons in the A–P axis. A consequence of this restricted cell body movement is that the position of medulla cell types along the A–P axis in the adult can act as a proxy for their developmental birthdate. Additionally, we find that the asymmetric A–P distribution of neuronal cell types results in unanticipated patterning of the retinotopic map. We show that, consistent with their localized cell body positions, the innervations of TmY17 neurons are restricted to the anterior medulla neuropil and the projections of TmY12 and Tm26 neurons are restricted to the posterior neuropil. Furthermore, in a syp^RNAi LOF background, extranumerary TmY17 cell bodies and their projections expand into more posterior medulla neuropil columns. NE temporal patterning therefore acts to both generate neural diversity in the medulla and, via the regulation of cell body positions along the A–P axis, pattern the retinotopic map. It is anticipated that this study may provide a mechanistic explanation for how stem cell patterning contributes to the formation of topographic maps in the vertebrate brain as well. In brain regions such as the olfactory epithelium and brainstem, the birthdate-dependent generation of neurons leads to the topographic mapping of their projections^63,64. Our findings in the medulla suggest that the temporal patterning of symmetrically dividing neural progenitors, combined with a directional progression of neurogenesis, could generate the birth order-dependent topographic maps observed in these vertebrate brain regions.

Methods

Fly strains

We used FlyBase (release FB2025_03) to find information on stocks, gene groups and protein function⁶⁵. Flies were reared at 25 °C on cornmeal food unless otherwise specified. OreR was used as the wild-type strain. The following fly stocks were used in this study: w¹¹¹⁸, vsx1-T2A-Gal4DBD (this study), w¹¹¹⁸;; hth-T2A-VP16AD (this study), w¹¹¹⁸;; GMR24F10-Gal4 (Dm4-Gal4, BDSC #49090), w¹¹¹⁸; GMR24F10-LexA (Dm4-LexA, BDSC #52696), MzVUM-Gal4⁶⁶, w¹¹¹⁸;; 13xLexAop2-myr::GFP (BDSC #32212), w¹¹¹⁸, hsFLP;; UAS-FRT > STOP > FRT-myr::smGFP-OLLAS, UAS-FRT > STOP > FRT-myr::smGFP-HA, UAS-FRT > STOP > FRT-myr::smGFP-V5, UAS-FRT > STOP > FRT-myr::smGFP-FLAG (MCFO-2, BDSC #64086), w¹¹¹⁸; UAS-FRT > STOP > FRT-myr::smGFP-V5, UAS-FRT > STOP > FRT-myr::smGFP-FLAG (BDSC #62124), w¹¹¹⁸, hsFLPG5::PEST (BDSC #62118), UAS-myr::GFP; Bl/CyO; Tm2/Tm6B (gift from Larry Zipursky), hs-Flp;; Act5C > CD2>Gal4, UAS-mRFP/ TM3, Sb¹ (gift from Dorothea Godt), w¹¹¹⁸;; pxb-Gal4 (gift from Claude Desplan), UAS-Flp; Gal80^ts; Act-FRT-stop-FRT-nls::βGal (gift from Claude Desplan), UAS-Flp; Act-FRT > STOP > FRT-myr::RFP; hh-Gal4 (gift from Claude Desplan), UAS-Flp; Optix-gal4; Act-FRT-stop-FRT-nls::βGal (gift from Claude Desplan), w¹¹¹⁸;; UAS-syp^RNAi (VDRC #33012), y¹sc*v¹sev²¹; UAS-eip93^RNAi (BDSC #57868), y¹sc*v¹sev²¹; UAS-mamo^RNAi (BDSC #44103), UAS-imp^RNAi; UAS-imp^RNAi/Tm6B (gift from Mubarak Hussain Syed), w¹¹¹⁸; VT048653-Gal4DBD (TmY15-Gal4DBD, BDSC #73733)⁴⁰, w¹¹¹⁸; R13E12-p65AD; R59C10-Gal4DBD (Tm3-Gal4, BDSC #86855), and w¹¹¹⁸;; Fru^NP21-Gal4 (Tm26-Gal4, BDSC #30027).

Immunohistochemistry

Immunohistochemistry analysis on larval and adult brains was performed as previously described⁶⁷. Immunohistochemistry analysis on pupal brains was performed as described for the adult. The following concentrations of primary antibodies were used to prepare the primary antibody solutions in PBT for a total volume of 100 μL: chicken anti-V5 (1:500; Abcam #ab9113), rat anti-FLAG (1:200; Novus Biologicals #NBP1-06712), rabbit anti-HA (1:500; Cell Signaling Technology #3724S), chicken anti-GFP (1:1000; Invitrogen #A10262), mouse anti-Brp (1:30; DSHB #nc82), rat anti-DE-cadherin (1:20; DSHB #DCAD2), rat anti-Chinmo⁶⁸ (1:500; gift from Nicholas Sokol), rat anti-DN-cadherin (1:20; DSHB #DN-Ex #8), guinea pig anti-Vsx1⁶⁶ (1:500), rabbit anti-Hth⁴² (1:1000; gift from Makoto Sato), mouse anti-Br-core (1:100; DSHB #Broad core (25E9.D7)), mouse anti-Armadillo (1:20; DSHB #N2 7A1 Armadillo), guinea pig anti-Mamo⁶⁹ (1:500; gift from Claude Desplan), mouse anti-Svp (1:200; DSHB #Seven-up 6F7), rabbit anti-E93⁷⁰ (1:300; gift from Daniel McKay), rabbit anti-Syp⁶⁹ (1:200; gift from Claude Desplan), rat anti-Imp⁶⁹ (1:200; gift from Claude Desplan), rabbit anti-Zfh1⁷¹ (1:2000; gift from Ruth Lehmann), rabbit anti-AstA (1:1000; Jena Bioscience #ABD-062), guinea pig anti-E93³⁶ (1:500; gift from Chris Doe), guinea pig anti-Chinmo⁶⁹ (1:200; gift from Claude Desplan), mouse anti-Chaoptin (1:20; DSHB #24B10), goat anti-βgal (1:1000; MP Biomedicals #0856028), rabbit anti-Ey³³ (1:250; gift from Claude Desplan), guinea pig anti-Slp2³³ (1:200; gift from Claude Desplan), mouse anti-Exd (1:200; DSHB #EXD B11M), guinea pig anti-Tll³³ (1:500; gift from Claude Desplan), guinea pig anti-Dpn³³ (1:1000; gift from Claude Desplan), rat anti-Dpn (1:50, Abcam #11D1BC7), guinea pig anti-Fru M (1:400; gift from Michael Perry) and rabbit anti-SoxN ⁷² (1:100; gift from Steven Russell).

Secondary antibodies were prepared at 1:500 in PBT for a total volume of 100 μL. Secondary antibodies were obtained from Invitrogen and Jackson ImmunoResearch Laboratories.

Generation of split-Gal4 lines

Vsx1 and Hth split-Gal4 lines were generated via CRISPR-mediated mutagenesis performed by WellGenetics Inc. The split-Gal4 cassettes containing T2A-VP16AD-RFP or T2A-Gal4DBD-RFP were knocked into the C-terminus of the vsx1 or hth coding region using homology-dependent repair, replacing the stop codon. The following gRNAs were used to target the C-terminus knock-in sites for each gene.

Hth:

> CTCGTAATCCGGCCCCGACC[CGG]

Vsx1:

>GAAGCAGGAGCAGGCGCATC[TGG] / ATCGATGTACATGTTTCATC[TGG]

Genetic techniques

MCFO clones

Clones of Vsx1–Hth neurons (Li2, TmY17, TmY15, TmY12, Tm23, Pm3a and Pm3b) were generated using the following split-Gal4 driver in combination with MCFO-2: w¹¹¹⁸, vsx1-T2A-Gal4DBD;; hth-T2A-VP16AD. Female adults were heat-shocked for 8 min at 37 °C and left to recover for 2 days before brain dissection.

Clones of wild-type TmY17 neurons were generated using the following split-Gal4 driver in combination with MCFO-2: w¹¹¹⁸;;TmY15-Gal4DBD/hth-T2A-VP16AD. Clones of TmY17 neurons in a syp^RNAi LOF background were generated using the following split-Gal4 driver in combination with MCFO-2: w¹¹¹⁸,vsx1-T2A-Gal4DBD;;hth-T2A-VP16AD/UAS-syp^RNAi. Female adults were heat-shocked for 8 min at 37 °C and left to recover for 2 days before brain dissection.

Clones of wild-type Tm26 neurons were generated using the following Gal4 driver in combination with MCFO-2: w¹¹¹⁸;;Tm26-Gal4. Male adults were heat-shocked for 8 min at 37 °C and left to recover for 2 days before brain dissection.

Loss-of-function analysis

To generate syp^RNAi LOF clones, hs-Flp;;Act5C > CD2>Gal4,UAS-mRFP/TM3,Sb¹ females were crossed to w¹¹¹⁸;;UAS-syp^RNAi males. Second instar larvae were heat-shocked at 37 °C for 8 min and dissected at the third instar larval stage or 5 h APF.

To knock down syp, e93, mamo and imp in Vsx1–Hth cells, males from each UAS-RNAi line (UAS-syp^RNAi, UAS-e93^RNAi, UAS-mamo^RNAi and UAS-imp^RNAi) were crossed to UAS-myr::GFP; Bl/Cyo; Tm2/Tm6B females. Male progeny were subsequently crossed to w¹¹¹⁸, vsx1-T2A-Gal4DBD;;hth-T2A-VP16AD females. Adult progeny of this cross were dissected and stained.

Developmental staging

To set up an egg lay, 40–50 adult flies were transferred to a bottle with yeast paste. Adult flies were allowed to lay for 3–4 h at 25 °C before being removed from the bottle. The bottle was subsequently incubated at 25 °C for 22 h. Following this incubation, all larvae that had hatched were removed from the bottle before placing it back at 25 °C. Newly hatched first instar larvae (0 h ALH) were collected from the bottle at 1 h intervals, placed into vials and left to develop at 25 °C until the desired developmental stage was reached.

Imp-Syp fluorescence intensity analysis

To quantify Imp and Syp levels, OreR larvae were developmentally staged and stained in a single immunohistochemistry batch. In each stain, Arm was used to label the OPC NE. All confocal microscopy images were acquired using the same acquisition settings, and ImageJ was used to make Imp and Syp fluorescence intensity measurements. In a single confocal slice, hand-drawn sections of the OPC NE (positive integrated density) or the empty background (negative integrated density) were selected on ImageJ using the ROI tool. For each channel, the negative integrated density was subtracted from the positive, resulting in the final integrated density. For each image, the relative Imp and Syp fluorescence intensity value was calculated by dividing the Imp or Syp integrated density by the Arm integrated density. These calculations were made for every timepoint except 48 h ALH. For 48 h ALH brains, the Arm integrated density was extrapolated. To do this, Arm intensities were averaged across the other time points (average Arm integrated density per unit area). Next, we calculated the average area of measurement for all 48 h ALH images and multiplied by the average Arm integrated density per unit area to generate the “48 h conversion for Arm normalization”. Finally, to normalize each Imp and Syp value to Arm, the relative Imp and Syp fluorescence intensity value was calculated by dividing the Imp or Syp integrated density by the “48 h conversion for Arm normalization” value.

Birthdating via EdU

Birthdating experiments within the larval feeding window (48–92 h ALH) were performed by administering a nucleotide analog, EdU (5-ethynyl-2’-deoxyuridine) to developing larvae. Larvae of the appropriate genotype were developmentally staged before being transferred to EdU-treated food.

48–92 h ALH EdU feeds

To make EdU-treated food, 10 mL of standard fly food was melted and poured into a vial. Next, 2 μL of EdU stock solution (50 mg/mL) was added to the food to achieve a final concentration of 10 μg/mL. EdU-treated food was stored at 4 °C for up to 4 days before use. Before transferring larvae to the EdU-treated food, refrigerated vials were left at room temperature for 1 h to warm. 48 h ALH larvae were transferred to the EdU food, and the vials were placed at 25 °C for the remainder of the experiment. Newly hatched adults were immediately transferred from the EdU-treated food to a standard food vial until they were dissected and stained.

6-hour EdU feeds

For shorter EdU feed experiments, 5 μL of EdU stock solution (50 mg/mL) was added to 10 mL of standard fly food to achieve a final concentration of 25 μg/mL. Red food dye was also added to the food to distinguish larvae that fed on the EdU-treated food from the others. Developmentally staged larvae were allowed to feed on the EdU-treated food for 6 h at 25 °C. After the feeding period, larvae with red dye visible in their midgut were transferred to a standard food vial for the remainder of development.

EdU cell counts

EdU-treated adult flies were dissected and stained using the previously described immunohistochemistry protocol⁶⁷ with one additional step. After secondary antibody incubation, brain tissues were washed in PBS overnight, followed by the Click-iT EdU reaction protocol (Thermo Fisher #C10338). For each of the 6 h EdU feed experiments, the total number of vsx1∩hth-Gal4 > GFP cells that had incorporated EdU was first counted in each optic lobe. Next, the neuron type identity of each EdU⁺ cell was identified using neuronal marker combinations to calculate the proportion of EdU⁺ neurons identified as a TmY17, TmY15, Tm23 or Pm3 within each optic lobe. To quantify the extent of EdU labeling in Dm4 and Tm26 neurons after a 60–92 h ALH EdU feed, the number of Dm4-Gal4 > GFP or Tm26-Gal4 > GFP cells was first counted in each optic lobe. Within each optic lobe, the proportion of neurons of each type that had incorporated EdU was then calculated.

Birthdating via pxb-Gal4 lineage trace

Neurons born after the larval feeding window (after 92 h ALH) were birthdated via heat shock-induced lineage trace experiments. w¹¹¹⁸;; pxb-Gal4 males were first crossed to UAS-Flp; Gal80^ts; Act-FRT-stop-FRT-nls::βGal females. The presence of a temperature-sensitive Gal80 (Gal80^ts) in the resulting progeny facilitated temporal control on UAS-Flp expression in the cOPC. Using this genetic scheme, all neurons generated from the cOPC post-memory trace induction are labeled by βGal.

Larval progeny of the above cross were developmentally staged, collected and raised at the restrictive temperature (18 °C). Slowed Drosophila development at 18 °C⁷³ was accounted for by stage larvae raised at the restrictive temperature (i.e. 92 h ALH at 25 °C corresponds to 184 h of development at 18 °C). Once at the appropriate stage, larvae were transferred to the permissive temperature of 29 °C for the remainder of development.

βGal cell counts

To quantify the extent of βGal labeling after memory trace induction, the total number of TmY17, TmY15, Tm23 or Pm2 cells was first counted in each optic lobe using neuronal marker combinations. The proportion of neurons of each type also labeled by βGal within each optic lobe was then calculated.

Imaging and image analysis

Images were acquired on a ZEISS LSM880 confocal microscope using a ×25 oil objective. Images were processed using Volocity imaging software or ImageJ.

Mapping the extent of TmY12, TmY17 and Tm26 medulla innervations along the D–V and A–P axes

Single-cell MCFO-labeled clones of wild-type Tm26, TmY12, TmY17 and syp^RNAi LOF TmY17 neurons in the adult optic lobe were generated as previously described. Following image acquisition, 3D rendering and measurement tools on Volocity imaging software were used to measure the A–P and D–V width of each adult medulla containing a neuronal clone using Brp or DN-Cad staining as a reference. Each medulla was then partitioned into equal-sized segments along the D–V axis and A–P axis to form a 10 × 10 coordinate grid. The spatial regions of the medulla innervated by each clone were then mapped to this coordinate grid. For a given cell type, the total number of clones innervating each spatial coordinate within the grid was counted. This data was prepared in the form of a heatmap using R. The intensity of color (white-to-red) in the heatmap reflects the proportion of clones of each cell type observed to innervate a given spatial coordinate.

Analysis of the published adult scRNA-seq dataset

A scRNA-seq dataset of the adult optic lobe (GEO accession: GSE142789)²⁴ was analyzed to identify vsx1 and hth co-expressing cell types. The pre-clustered Seurat object (Adult.rds, GSE142787) was analyzed using the assigned cluster identities located in the “FinalIdents” field of the metadata. Clusters with significantly upregulated vsx1 or hth expression (Padj < 0.05, Wilcoxon rank sum test) were identified using the FindAllMarkers function in Seurat 3.1.5 under the following parameters: logfc.threshold = 0, min.pct = 0.3.

Cells within clusters identified to co-express vsx1 and hth were subsetted from the dataset and re-clustered relative to one another using Seurat 3.1.5. All clustering functions were run under default parameters and using 100 principal components (PCs). Downstream analysis was performed on the re-clustered cells defined by FindClusters at a resolution of 0.8. The FindAllMarkers function was run on the re-clustered data using default parameters to identify unique combinations of upregulated marker genes for each of the cell clusters.

Statistics and reproducibility

For the expression patterns shown in Figs. 1b–h, 2d–m, 3a–c, e–n, 4a–l, n–s, u–x, 5a–l, n, q, s,v, 6e–m, 7a–d and Supplementary Figs. 1a–i, 2i–s, 3b, c, 4a–j, 5a, b, d, e, g, h, j–o, 6a–f, 7a–h at least five brains were imaged for each experiment. For experiments accompanied by quantitative analyses, the numbers of optic lobes/neurons/clones analyzed are indicated in the corresponding figure captions. Sample sizes were determined according to standards in the field.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.