Mutations in abnormal spindle disrupt temporal transcription factor expression and trigger immune responses in the Drosophila brain

Abstract

The coordination of cellular behaviors during neurodevelopment is critical for determining the form, function, and size of the central nervous system (CNS). Mutations in the vertebrate Abnormal Spindle-Like, Microcephaly Associated (ASPM) gene and its Drosophila melanogaster ortholog abnormal spindle (asp) lead to microcephaly (MCPH), a reduction in overall brain size whose etiology remains poorly defined. Here, we provide the neurodevelopmental transcriptional landscape for a Drosophila model for autosomal recessive primary microcephaly-5 (MCPH5) and extend our findings into the functional realm to identify the key cellular mechanisms responsible for Asp-dependent brain growth and development. We identify multiple transcriptomic signatures, including new patterns of coexpressed genes in the developing CNS. Defects in optic lobe neurogenesis were detected in larval brains through downregulation of temporal transcription factors (tTFs) and Notch signaling targets, which correlated with a significant reduction in brain size and total cell numbers during the neurogenic window of development. We also found inflammation as a hallmark of asp mutant brains, detectable throughout every stage of CNS development, which also contributes to the brain size phenotype. Finally, we show that apoptosis is not a primary driver of the asp mutant brain phenotypes, further highlighting an intrinsic Asp-dependent neurogenesis promotion mechanism that is independent of cell death. Collectively, our results suggest that the etiology of the asp mutant brain phenotype is complex and that a comprehensive view of the cellular basis of the disorder requires an understanding of how multiple pathway inputs collectively determine tissue size and architecture.

Autosomal recessive primary microcephaly (MCPH) is a neurodevelopmental disorder characterized by a reduction in brain size, intellect, and life span whose etiology remains poorly understood. Here, Mannino et al. provide the neurodevelopmental transcriptional landscape of a Drosophila model of MCPH5 caused by mutations in abnormal spindle. They characterize asp expression across cell-types and developmental stages, solidifying its role as a neurogenesis-promoting factor, and identify new pathways important for Asp-dependent brain growth through gene expression profiling and mutational analysis.

Introduction

Autosomal recessive primary microcephaly (MCPH) is a congenital neurodevelopmental disorder characterized by a reduction in overall brain size, intellectual disabilities, and shortened life span (Faheem et al. 2015). While the clinical manifestations of the disorder are well-characterized, the underlying molecular mechanisms responsible have been difficult to pinpoint. This is due to the increasing number of unique genes (>30) that have been found mutated in human MCPH patients, each of which has known roles in diverse cellular functions such as mitosis, centriole structure and function, DNA damage and repair, chromatin remodeling, and nuclear envelope integrity (Siskos et al. 2021). This genetic diversity implies that multiple cellular pathways can give rise to the small brain phenotype, although our understanding of how this occurs remains limited.

Abnormal Spindle-Like, Microcephaly Associated (ASPM) gene is the most commonly mutated gene found in human MCPH patients. Designated as MCPH5 (OMIM #60871), it accounts for more than 50% of all reported MCPH cases (Muhammad et al. 2009). The Drosophila melanogaster ortholog, abnormal spindle (asp), was identified before its human counterpart and was originally described as an essential mitosis gene, required for proper mitotic spindle morphology, mitotic progression, and chromosome segregation (Ripoll et al. 1985). Cell biology, genetic, and biochemical studies by multiple labs have since provided a molecular explanation for how Asp functions as a “glue” to maintain spindle pole and centrosome-pole cohesion during metaphase by associating with microtubule minus ends, as well as having additional roles in actin cytoskeleton regulation that may be important for proper tissue architecture (Saunders et al. 1997; do Carmo Avides and Glover 1999; Wakefield et al. 2001; Rujano et al. 2013; Ito and Goshima 2015; Schoborg et al. 2015).

The observed defects in mitosis provide an intuitive hypothesis for how mutations in asp/ASPM lead to reduced brain size, which can be broadly summarized into a defective neurogenesis program (O’Neill et al. 2018). Neurogenesis is the developmental process by which neural stem cells (NSCs) produce a final pool of fully differentiated, postmitotic neurons and glia in the central nervous system (CNS) (Paridaen and Huttner 2014). In primates, brain size follows linear cell scaling rules with cell number being the primary driver (Herculano-Houzel et al. 2007). Thus, mechanisms that promote efficient neurogenesis such as error-free mitosis are thought to have a significant influence on total neuron and glia numbers and ultimately final brain size.

Given Asp/ASPM's prominent role in mitosis, much of the earlier work attempting to link the cell biology to MCPH focused on cell division defects in the etiology of the disorder. In mice, Aspm was shown to regulate mitotic spindle orientation, preventing a premature switching of proliferative symmetric divisions in neuroepithelial cells (NECs) to neurogenic, asymmetric divisions to ensure the appropriate amount of cortex neurons and glia (Fish et al. 2006). A similar defect in spindle orientation was also found in human cultured cells lacking ASPM, along with defects in cytokinesis (Higgins et al. 2010). Drosophila asp has additional roles in spindle pole focusing and centrosome-pole attachment, which led to delayed anaphase onset, extended mitosis, and errors in chromosome segregation, which were thought to be responsible for the small brain phenotype in flies (Rujano et al. 2013; Ito and Goshima 2015; Schoborg et al. 2015).

However, disrupted mitotic spindle morphology alone is not sufficient to explain the small brain phenotype in flies, as only the N-terminal half of the Asp protein was sufficient to rescue brain size phenotypes despite the fact that mitotic spindle morphology was still disrupted in these animals (Rujano et al. 2013; Schoborg et al. 2015, 2019). Furthermore, recent vertebrate studies have suggested that Aspm has additional, nonmitotic functions that may be the primary driver of the small brain phenotype (Capecchi and Pozner 2015; Jayaraman et al. 2016), suggesting that our knowledge of the cellular basis of asp/ASPM MCPH is incomplete and likely involves a combination of both mitotic and interphase roles that contribute to the small brain phenotype.

Next-generation sequencing and other omic-based approaches can provide insight into the spectrum of biological pathways contributing to MCPH (Wang and Wang 2019), yet they have been underutilized in the MCPH field owing to limitations in model systems and the difficulty associated with acquiring viable samples from human patients (Yang et al. 2012; Evrony et al. 2017; Johnson et al. 2018; Wang et al. 2020). Drosophila provides a unique model to explore these pathways in a temporal fashion, as the bulk of neurogenesis, neuronal remodeling, and maintenance of the final CNS development program occur at distinct developmental stages (Truman and Bate 1988), thus allowing for identification of the relevant cellular pathways involved and also when in the neurodevelopmental pipeline they become important for brain size determination. Furthermore, the fly genetic tools available allow researchers to extend their analysis beyond discovery and into functional testing of hypotheses based on high-throughput data to better understand how multiple inputs are coordinated to affect CNS growth and development.

Here, we describe the developmental cell expression pattern of asp in Drosophila brains and identify most of the cell types that are responsible for the small brain phenotype in asp mutant animals. To further explore the mechanism of asp-dependent brain growth and development, we performed transcriptome analysis for asp mutant brains during the larval, pupal, and adult stages of neurodevelopment. Using coexpression and functional network analysis, we identify multiple biological pathways and gene regulatory modules enriched in the data sets. We found a number of temporal transcription factors (tTFs) that are critical for optic lobe neurogenesis and neuronal diversity that were downregulated in larval brains, along with Notch signaling targets. These genes have known roles in the neurogenic cell types we identified that require asp expression for correct brain size. Disruption of these key neurogenic factors also correlated with a significant reduction in asp mutant brain size and CNS cell number during the larval stage of development, providing a plausible mechanism to link neurogenic defects to the small brain phenotype. We also identify inflammatory signatures during all developmental stages, suggesting that immune system activation may be a hallmark of asp mutant brains. Genetic analysis revealed that this partly contributes to the size phenotype as well. Other signatures included proteolysis, stress, metabolic, and actin cytoskeleton-related pathways that were enriched in at least 2 of the 3 stages. Surprisingly, we did not find apoptosis to be a primary driver of the asp mutant brain phenotype, lending further support to a model in which asp promotes proper neurogenesis and cell number through cell death–independent mechanisms. Together, our results suggest that multiple inputs contribute to the asp brain phenotype and provide a platform for future functional studies to explore the relative contributions of these pathways in the etiology of MCPH5.

Materials and methods

Fly stocks and husbandry

All stocks and crosses were maintained on standard cornmeal-agar media at 25°C. The asp mutant alleles (asp^T25 and asp^Df) and asp transgenic rescue lines were previously described (Schoborg et al. 2015, 2019). The following lines were obtained from the Bloomington Stock Center: w¹¹¹⁸; Rel^E38, e[s] (BS#: 9458); P{ry^+t7.2 = Dipt2.2-lacZ}1, P{w^+mC = Drs-GFP.JM804}1, y¹ w*; Dif¹  cn¹  bw¹(BS#: 36559); w¹¹¹⁸; P{w^+m* = GAL4}repo/TM3, Sb¹ (BS#: 7415); w[*]; P{w[ + mW.hs] = GawB}insc[Mz1407] (BS#: 8751); P{w[ + mC] = GAL4-elav.L}2/CyO (BS#: 8765); y[1] w[*]; P{Act5C-GAL4-w}E1/CyO (BS#: 25374); and w[*]; P{w[ + mC] = UAS-p35.H}BH1 (BS#: 5072). The Dif RNAi line was obtained from the Vienna Drosophila Resource Center (#100537). The UAS-H2Av::RFP line was obtained from Laura Buttitta. Mutations were verified using single-wing PCR (Carvalho et al. 2009) and Sanger sequencing. All UAS and Ubi transgenic lines were generated in the yw mutant background by BestGene (Chino Hills, CA, USA).

Developmental timing and brain dissection

Animals were staged based on the following criteria: wandering third instar larvae were collected from noncrowded vials once they emerged from the food; P7 pupae were collected as white prepupae, placed in a 35-mm dish with a moist Kim wipe, and allowed to develop for an additional 48 h at 25°C; adult females were collected shortly after eclosion and aged for 3–5 days. Larval and pupal brains were rapidly dissected in fresh, room temperature (RT) SF900 media. Adults were dipped into 100% EtOH for 3–5 s prior to dissection in fresh SF900 to remove the waxy covering of the cuticle and allow for easier dissection. Dissected brains were maintained in fresh SF900 media for no longer than 15 min at RT. Four biological replicates consisting of 10–15 brains for each genotype and developmental stage were used for downstream processing.

Total RNA extraction

Total RNA was extracted using Trizol reagent (Invitrogen) following the manufacturer's instructions. The RNA was resuspended in 12.5 µL of sterile Milli-Q water and treated with Turbo DNAse (Invitrogen) to eliminate DNA contamination. Quantity and quality of RNA were assessed using a Nanodrop 2000 spectrophotometer (Thermo Scientific, USA) and 1% (w/v) agarose gel electrophoresis, respectively.

RNA-seq and DGE analysis

Library preparation and RNA-sequencing (RNA-seq) was performed by Novagene (UC, Davis, CA, USA) using an Illumina NovaSeq 6000 platform. All samples had a RNA Integrity Number (RIN) score of ≥6.5. Sequencing generated a total of 881 million paired-end reads of 150 bp. Raw sequencing reads were trimmed and filtered using FastQC and Cutadapt with a length cutoff of 20 nt (Andrews 2010; Martin 2011). Surviving reads were mapped to the D. melanogaster reference genome (dm6) using the Rsubread package (Liao et al. 2019). Gene expression levels were estimated using featureCounts software with default settings (Liao et al. 2014). Differential gene expression (DGE) analysis was performed using the edgeR Bioconductor package, filtering for Counts per Million (CPM) < 1 (Robinson et al. 2010). Genes with a logF.C. of ≥0.5/≤−0.5 and an adjusted P-value of ≤0.05 were assigned as differentially expressed.

Functional annotation and network analysis

Identification of biologically relevant pathways and gene sets was performed using DAVID (v2021) and the EnrichmentMap and GeneMANIA plugins for Cytoscape (v3.9.1) (Shannon et al. 2003; Huang et al. 2009; Merico et al. 2010; Warde-Farley et al. 2010; Sherman et al. 2022). Differentially expressed genes (DEGs) were first analyzed in DAVID using the Functional Annotation Tool. Parameters included (1) functional annotations: UP KW biological process, UP KW cellular component, and UP KW molecular function; (2) gene ontology (GO): GOTERM BP direct, GOTERM CC direct, and GOTERM minimal fragment (MF) direct; and (3) pathways: KEGG pathway. Functional annotation clustering was then performed using default settings and a medium classification stringency. Functional annotation charts were also generated in DAVID using default settings, which served as the input for EnrichmentMap. Cutoff values using the overlap index test for each data set included P-value of 0.05, False Discovery Rate (FDR) Q-value of 0.1, and overlap of 0.6. Gene set nodes were then analyzed in GeneMANIA using default settings to the visualize connectivity between genes based on previously published coexpression, predicted, colocalization, physical interactions, and genetic interactions data sets.

ICA

A second analysis pipeline was used to assess the functional enrichment of DEGs based on Drosophila-specific gene expression signatures (Rusan et al. 2020). These gene expression signatures were identified through independent component analysis (ICA) of 3,346 microarray data sets generated by the fly community and consist of 850 modules of coregulated genes. ICA modules were encoded as a sparse nonnegative 18,952 microarray probe sets by 850 modules weight matrix derived from an 18,952 microarray probe sets by 425 independent components ICA S matrix (see Rusan et al. 2020 for details). A total of 850 sets of the most heavily weighted coregulated genes were created by discretizing (probe set weight ≥ 3) each of the ICA module matrix vectors and used to determine module GO term and KEGG pathway enrichments (Supplementary File 4). To project the expression data in this report onto the ICA module matrix, the matrix probe set features were first converted to FBgn features compatible with edgeR logF.C. vectors using the biomaRt package. ICA module matrix weights for the relatively few probe sets that mapped to the same FBgn were averaged to yield the 13,801 FBgn by 850 modules ICA module matrix used here. Prior to projection onto the FBgn-based ICA module matrix, edgeR outputted genotype pair FBgn logF.C.s vectors were mean 0 centered and variance 1 scaled. After restricting the ICA module matrix and logF.C. vectors to the FBgn features shared between them, scalar projection of the logF.C. vectors onto the ICA module matrix was performed as described previously (Rusan et al. 2020). In addition, 100 random permutations of each genotype pair logF.C. vector (100 shuffles of FBgn label to logF.C. value relationships) were also projected onto the ICA module matrix. The mean and SD of random permutation scalars per module were used to compute enrichment Z-score and P-value significance values for each module in a genotype pair logF.C. vector. A module was considered significantly enriched if it obtained a Z-score of ≥+3/≤−3 and a P-value of ≤0.05.

Gene expression: RT-qPCR

cDNA was synthesized via the iScript cDNA Synthesis Kit (BioRad, USA) using 1 μg of total RNA. The resultant cDNA was then used 1:10 in the qPCR reaction, which consisted of iQ SYBR Green Supermix (BioRad) and gene-specific primers. Primers are available upon request. The following cycling parameters were used: 95°C for 5 min, 40 cycles of 95°C for 10 s, and 60°C for 45 s ending with a melting curve. Relative gene expression was analyzed by the multiple reference gene method. Elongation factor 1-alpha (ef1-α) and RP49 were used as the internal reference genes. Relative quantities and data normalization followed the formulas detailed in Hellemans et al. (2007). The comparative Cq (ΔΔCq) method was employed to calculate the relative expression ratios (RERs). Three technical replicates were performed for each of the 4 independent biological replicates assayed. Statistical analysis was performed using ANOVA, Bonferroni's posttest, and t-test when applicable using PRISM 9 (GraphPad Software, USA).

Immunohistochemistry and antibodies

Larval and adult brains were carefully dissected in SF900 media, transferred to 1.5-mL tubes containing 1.2% paraformaldehyde (PFA) diluted in 1 × PBS, and fixed for 24 h while nutating at 4°C. Fixed brains were then washed in 0.5% Triton X-100 in 1x PBS (PBT) for 3 × 15 min at RT, blocked for 1.5 h in 0.5% PBT with 5% normal goat serum, and nutated with the following antibodies: anti-Bruchpilot antibody (1:30, nc82, DSHB), anti-cleaved death caspase-1 (DCP-1, 1:100, Cell Signaling Technology), anti-β-tubulin (1:250, E7, DSHB), anti-phospho-histone H3 (Ser10) (1:1,000, Millipore Sigma), anti-discs large (1:50, 4F3, DSHB), anti-deadpan (1:500, 11D1BC7, Abcam), anti-Elav (1:100, 9F8A9, DSHB), anti-Repo (1:30, 8D12, DSHB), anti-lethal of scute (1:5,000, gift from Andrea Brand), anti-GFP (1:500, AB290, Abcam), and anti-Prospero (1:100, MR1A, DSHB) for 4 h at RT before transferring to 4°C for 2 overnights. The brains were then washed using 0.5% PBT (4 × 15 min) and incubated 1:500 with secondary antibody (anti-mouse/rabbit Alexa-488/647, Invitrogen/Thermo Fisher) for 4 h at RT before transferring to 4°C for 3 overnights. Brains were again washed using 0.5% PBT (4 × 15 min) and stained with DAPI (0.1 µg/mL). Following a postfixation step with 4% PFA at RT for 4 h, the brains were washed using 0.5% PBT (4 × 15 min) and mounted on poly-L-lysine coverslips. The coverslips were transferred through an ethanol dehydration series (30%, 50%, 75%, 95%, and 100% ×3), cleared using xylene, and mounted in DPX mounting media. Slides were allowed to cure for at least 2 overnights before imaging.

Microscopy

All imaging was conducted on an Olympus IX83 microscope fitted with a Yokogawa CSU-W1 Dual Disk SoRa, dual Hamamatsu Orca Flash 4.0 V3 sCMOS cameras, and Plan S-Apo 40× and 100× Si Oil Objectives (NA 1.25 and 1.35) operated by cellSens software.

Image analysis

For the image pixel intensity counts for comparing Asp-fTRG signal in larval, pupal, and adult brains (Supplementary Fig. 1d and e), all brains were dissected and labeled with α-GFP. One larval brain was used to establish the imaging parameters (laser power and exposure time), and then at least 5 brains from each developmental stage were acquired with these same parameters. Following background subtraction, we quantitated the mean pixel intensity for each region of interest (ROI) consisting of a cell body located in the top third of the z-stack using FIJI. All mean pixel intensity values were normalized to the adult, which was set to 1.

For the apoptotic cell counts, DCP-1-positive cell counts were obtained using the surface counter feature in Imaris (v9.9.1). Uniform threshold values were established for the intensity of cells positive for DCP-1 based on positive control and negative control brains. Other parameters were established prior to analysis including cell seed diameter for completing a morphological split of touching or grouped cells, surface smoothing, quality evaluation of the detected surfaces, and the minimum number of voxels per object. All raw cell counts were further normalized by dividing the cell count for the optic lobe by the total volume of the optic lobe, which was calculated using the segmentation feature in Dragonfly software (Object Research Systems, v2020.2.0.941). Thresholding was utilized during segmentation in Dragonfly to ensure measurements only included the optic lobe of interest.

µ-CT

Sample preparation, imaging, and analysis were carried out as previously described (Schoborg et al. 2019; Schoborg 2020). I2KI was used as the contrast agent. Samples were scanned using a SkyScan 1172 desktop scanner controlled by SkyScan software (Bruker). X-ray source voltage and current settings: 40 kV, 250 μA, and 4 W of power. A cooled 14-bit 11-Mp (402 × 2,688) CCD detector coupled to a scintillator was used to collect X-rays converted to photons. Medium camera settings at an image pixel size of 2.95 μm were used for fast scans (∼20 min), which consisted of about 300 projection images. Frame averaging used was 2. Tomograms were generated using NRecon software (Bruker MicroCT, v1.7.0.4). Dragonfly (Object Research Systems, v2020.2.0.941) was used for manual brain segmentation and determination of brain and optic lobe volume.

Flow cytometry

We followed the method of Nandakumar et al. (2020) to obtain total cell counts from adult brains. Single adult brains were dissected in SF900 media and transferred to a sample tube containing 90-μL trypsin–EDTA, 10-μL 10× PBS, and 2 μL of Vybrant DyeCycle Violet stain (Thermo Fisher Scientific). Brains were incubated for 20 min at RT before being triturated using a low retention P200 pipette tip for 90 s. Four hundred microliters of the trypsin–EDTA–DyeCycle solution was then added to allow for additional digestion of the tissue for 45 min at RT. The sample was then diluted with 500 μL of 1× PBS and gently vortex to resuspend the cells. Samples were run at a slow flow rate (∼280 events/s or 1 mL/15 min) using a BD FACSMelody (BD Biosciences) until the entire volume of the tube was empty. Data were analyzed using FlowJo software (v10.7.2, BD Biosciences). Events were gated using relative size (FSC) and granularity (SSC). To establish fluorescent vs nonfluorescent populations of events, unlabeled Oregon-R brains were used to establish the maximum nonfluorescence threshold based on intensity. Events that populated above this threshold in the DyeCycle-labeled samples were gated as positive and counted. We verified that most of these events consisted of brain cells by subsequent flow sorting and immunofluorescence microscopy using an anti-lamin antibody (1:100, ADL101, DSHB) to visualize the nuclear envelope. We also verified this population using a UAS-H2Av::RFP line driven by Actin-Gal4, which were double positive for both DyeCycle Violet and RFP signal indicating intact fluorescent nuclei (Supplementary Fig. 2).

Results

Asp expression in the Drosophila brain is restricted to the neurogenic larval phase of development and is found in all major NSC populations

Despite decades of Asp studies in Drosophila, we lack detailed knowledge of its cell and developmental expression behavior in most tissues, including the brain. Fly brain development begins with a neurogenesis phase during the embryonic and larval stages, where symmetrically dividing NECs delaminate and form the asymmetrically dividing neuroblasts (NBs). NBs self-renew and generate ganglion mother cells (GMCs) that will later produce neurons and glia. Embryonic neurogenesis accounts for 10% of the neurons and glia that will contribute to the adult CNS, whereas 90% are generated during the larval period following a brief NB quiescence state (Homem and Knoblich 2012). Remodeling of the CNS occurs primarily during the pupal stage, and the end result is an adult brain consisting of ∼200,000 cells. This represents a 1,900% increase in cell number compared to the ∼10,000 cells found in the first instar larval brain (Truman and Bate 1988; Campos-Ortega 1993) (Fig. 1a and b).

Fig. 1.

Asp is highly expressed in mitotic cells of the larval brain. a) Summary of fly neurodevelopment. Neurogenesis begins during the embryonic (Em) stage, where NECs delaminate from the neuroectoderm and form asymmetrically dividing NBs. NBs divide in a self-renewing fashion to generate GMCs, which then further divide into postmitotic neurons and glia. A brief quiescent period (Q) precedes the larval neurogenesis period, with 90% of the cells present in the adult CNS generated during the L1, L2, and L3 stages. Neuronal remodeling occurs during the pupal stages, shaping the adult CNS into its final form consisting of ∼200,000 cells. b) Cartoon representation of the various mitotic cell populations in the third instar larval brain that make neurons and glia (omitted for clarity). See text for description of each. c) Graph of asp mRNA expression levels throughout neurodevelopment in the CNS or entire head from L3 larva, P8 pupa, day 1 (1D) and day 20 (20D) male (M) and female (F) adults. Expression values shown are normalized reads Per kilobase per million (RPKM), acquired from FlyBase release FB2023_03. d) An entire L3 larval optic lobe stained with α-GFP to visualize Asp-fTRG protein and discs large (α-Dlg) to label cell outlines. Three different Z-planes are shown to highlight the relevant NSC populations: mNBs, cNBs; the LPCs and NECs of the OPC; and both NEC populations, the OPC and IPC. Asp is highly enriched in all dividing cell types. e) A cluster of cells consisting of a cNB marked with Deadpan (α-Dpn) (white outline), the GMCs (cyan outline) and neurons (yellow outline) marked by Prospero (α-Pros). Though Prospero labels both GMCs and neurons in the larval brain (Carney et al. 2013), the GMCs can be identified through stronger Asp-GFP signal compared to the neurons and also by evaluating the pattern of α-elav staining, which only labels neurons (f, yellow outline). Two different Z-planes are shown in f) to highlight weak but detectable Asp signal in both younger (Z #1) and older (Z #2) neurons. Scale Bars: 10 μm d); 3 μm e); and 5 μm f). Panels a) and b) were created with BioRender.com.

Using publicly available transcriptomic data from the modENCODE Anatomy RNA-Seq project at FlyBase (Gramates et al. 2022; Brown 2014), we found that asp mRNA expression in the brain occurs exclusively during the larval stage of development, with no transcripts detected in mid-pupal (P8) or adult heads (Fig. 1c). To verify protein levels, we used an Asp reporter line derived from the Tagged FlyFos TransgeneOme (fTRG) library (Sarov et al. 2016). This construct consists of a C-terminally GFP tagged asp genomic fosmid clone with all endogenous regulatory elements (hereafter referred to as Asp-fTRG; Supplementary Fig. 1a and b). The subcellular localization of Asp-fTRG in dividing NBs was identical to previous reports with strong localization to astral microtubules during prophase and the minus end of spindle microtubules during metaphase (Supplementary Fig. 1c and h) (do Carmo Avides and Glover 1999; Schoborg et al. 2015). At the tissue level, we detected a strong Asp protein signal in specific regions of the larval brain. We also noted a strong reduction in Asp protein signal in the brain as animals progressed through the pupal and adult stages, with the remaining signal weakly visible in the cell bodies of the neurons surrounding the neuropil (Supplementary Fig. 1d and e). This restricted developmental expression pattern suggests that Asp is primarily needed during a specific window of CNS development. In the case of flies, this corresponds to the neurogenic phase of development, which is when postmitotic neurons and glia are being produced from an actively dividing population of NSCs (Fig. 1a and b).

To further explore the cell-type–specific expression pattern of Asp in the larval brain, we performed immunostaining in our Asp-fTRG reporter line with well-characterized markers. We found that Asp is highly expressed in all mitotic populations of cells (Fig. 1b). This includes the symmetrically dividing NECs, which are organized into 2 distinct layers referred to as the outer proliferation center (OPC) and the inner proliferation center (IPC) that can be identified via discs large (α-dlg) staining (Fig. 1d). Asp was also highly expressed in each asymmetrically dividing NB population [medulla NBs (mNBs) and type I/II central brain NBs (cNBs)], as visualized using both α-dlg and the NBs marker deadpan (α-dpn) (Fig. 1d–f; Supplementary Fig. 1f). Additionally, we found Asp expression in the lethal of scute (l′sc)-positive transition zone (TZ) cells, which are undergoing the NEC to mNB fate switch (Supplementary Fig. 1h). Strong Asp signal was also observed in the lamina precursor cells (LPCs), which divide to make the neurons and glia of the lamina neuropil and are located laterally to the OPC (Fig. 1d). Finally, we also observed strong Asp signal in all Prospero (α-pros)-positive GMC populations that arise from the asymmetric division of both cNB and mNB (Fig. 1e and f, cyan outlined population). Together, these data suggest that strong Asp expression is a hallmark of all actively dividing NSC populations of the larval brain.

We next asked whether Asp protein was detectable in postmitotic neurons and glia of the larval brain. We found no evidence of Asp protein in repo-labeled glial cells (Supplementary Fig. 1g). However, we did observe a weak but detectable signal in neurons (Fig. 1e and f, yellow outlined population). The weak signal was primarily cytoplasmic and found only in the cell bodies. These data agree with the weak cell body signal we observed in pupal and adult brains (Supplementary Fig. 1d). Given that no asp mRNA transcripts are made during these stages (Fig. 1c), we believe that the weak neuronal signal at all developmental stages is likely the result of cytoplasmic inheritance from the GMC during the terminal division, rather than active transcription and translation to make new Asp protein in neurons.

Asp mutations affect the neurogenic phase of development and the small brain phenotype correlates with a reduction in total brain cell number

We previously reported that asp mutant adult brain volume is significantly reduced compared to wild-type (WT) animals, with the optic lobes showing the greatest reduction in size based on microcomputed tomography (μ-CT) analysis (Schoborg et al. 2019). In this study, we also wanted to extend our volume analysis to the larval and pupal stages to better understand the onset of the volume decrease phenotype and highlight the neurodevelopmental time course.

We chose 3 developmental time points to assess: (1) the late third instar larvae, when NSCs are actively dividing to generate neurons and glia (neurogenesis) (Truman and Bate 1988); (2) the pupal brain at ∼48 h after pupal formation (APF) when neuronal remodeling is occurring (Meltzer and Schuldiner 2022); (3) the adult brain 3–5 days after eclosion when the majority of neurogenesis and remodeling have completed and thus reflect the end-point of neuronal development (Li and Hidalgo 2020). Both male and female asp mutants show a significant percent decrease in entire brain size compared to asp heterozygous control animals (asp^T25/+, hereafter referred to as WT; Schoborg et al. 2019) (Supplementary Fig. 2b). However, viable asp mutant males are difficult to obtain due to significant lethality; thus, we used females for the rest of our analyses.

We found a 33 and 28% decrease in the entire CNS in asp mutants compared to WT in third instar larval and pupal brains, respectively (Fig. 2a; Supplementary Fig. 2a, c, and d). We also found a 23% decrease in adult asp entire brain size that was highly significant (P < 0.0001) (Fig. 2a; Supplementary Fig. 2f). A similar, but more pronounced trend, was also observed when measuring optic lobe volume alone, with a 58 and 42% reduction in size seen in pupal and adult brains, respectively (Supplementary Fig. 2e and g). Given that the size decrease is readily evident toward the end of the neurogenic period (third instar larva) and does not decrease further at later developmental stages, these results suggest that the asp-induced small brain phenotype results primarily from neurogenic defects, analogous to its mammalian ASPM ortholog (Fish et al. 2006).

Fig. 2.

Asp mutant brains are significantly reduced in size and total cell number, which can be mostly restored by the asp rescue fragment. a) μ-CT volume measurements from WT and asp mutant larva, pupa, and adults of the entire brain. b) Volume measurements of the entire brain from asp rescue and rescue control animals at the larval and adult stages. c) Validation of the flow cytometry method used to count neuronal cells from single brains. Animals expressing an RFP-tagged H2Av protein under control of the Act-Gal4 driver were colabeled with DyeCycle Violet. Events displaying high signal for both channels (V+ and R+) were taken as the intact cell nuclei population and used to generate the counts displayed in d) larval WT vs asp mutant, e) adult WT vs asp mutant, f) larval control and asp rescue, and g) adult control and asp rescue brains. Each dot represents the count from a single brain (n = 10). h) Metaphase mitotic spindles of larval NECs undergoing symmetric division, labeled with anti-β-tubulin to visualize microtubules and anti-pH3 to visualize mitotic chromosomes in heterozygous control (asp^T25/+), asp mutant (asp^T25/asp^Df), and asp rescue (ubi-GFP::asp^MF/+; asp^T25/asp^Df) brains. Yellow arrowhead denotes unfocused spindle poles, and red arrowhead indicates partially detached centrosomes. i) Genetic rescue experiments using the UAS/Gal4 system with the asp rescue fragment. Insc-Gal4, Elav-Gal4, and Repo-Gal4 were used to express the fragment in most NBs, neurons, and glia, respectively. Entire brain measurements are shown. j) Same genetic rescue assay outlined in i) but measuring optic lobe volume only for the indicated drivers. For all μ-CT measurements, data are represented as the T-ratio, which normalizes optic lobe volume to thorax width (body size) (Schoborg et al. 2019). Heterozygous control (asp^T25/+) was set to 1, and the subsequent genotypes were normalized accordingly. n ≥ 5 brains, Welch's t-test. ns, P > 0.05; *P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001; ****P ≤ 0.0001. Error bars represent standard deviation. Scale bars: 5 μm.

We also measured larval and adult brains from our previously described asp rescue strain (Schoborg et al. 2019). The asp rescue strain expresses an N-terminally tagged GFP Minimal Fragment (MF) of Asp consisting of the first 597 amino acids of the protein in the asp mutant background. Mitotic spindle morphology is still disrupted in asp rescue animals, with unfocused spindle poles and detached centrosomes evident, much like the asp mutants (Fig. 2h). However, this asp rescue fragment is able to restore both the entire brain and optic lobe sizes >90% in asp mutant adult brains, as well as neuropil morphology (Fig. 2b; Supplementary Fig. 2i and j) (Schoborg et al. 2019). A similar trend (>88% rescue) was observed in the larval brain measurements as well (Fig. 2b; Supplementary Fig. 2h). These results suggest that the Asp^MF rescue fragment is largely capable of performing the necessary cellular roles of the full-length protein that are required for proper brain size specification during the neurogenic window of development, which is independent of Asp's role in mitotic spindle morphology.

To further investigate the cellular defects possibly contributing to the asp mutant small brain phenotype, we performed flow cytometry on single fly brains (larval and adult) to determine total neuronal cell number. In primates, neuronal cell number is the primary driver of overall brain size (Herculano-Houzel et al. 2007). Utilizing a fly brain–specific workflow outlined by the Buttitta Lab (Nandakumar et al. 2020), we validated its use in our hands and identified the relevant intact cell populations (Fig. 2c; Supplementary Fig. 3). We found a highly significant difference in cell number, with a 42 and 38% decrease in asp mutant third instar larva and adult brains, respectively, compared to WT (Fig. 2d and e). We also measured neuronal cell number in asp rescue animals and found a similar trend to that observed in our μ-CT measurements of overall brain size: only an 18 and 6% decrease in cell number was observed in larval and adult asp rescue brains compared to asp rescue controls, respectively (Fig. 2f and g). Together, these results confirm that asp mutant brains have reduced numbers of cells that can largely be restored by the asp rescue fragment and that this reduction in cell number may be the primary driver of brain size in Drosophila, similar to primates (Herculano-Houzel et al. 2007).

Asp expression is required in actively dividing cells to promote proper brain size

We next wanted to test whether Asp expression in specific cell types could rescue the small brain phenotype in asp mutant animals. We used the UAS-Gal4 system to express the asp rescue fragment (UAS-GFP::Asp^MF) transgene in the asp mutant background (Brand and Perrimon 1993). We used 3 well-characterized Gal4 drivers whose expression patterns were verfieid using the G-TRACE reporter system (Evans et al. 2009): Insc-Gal4 (insc^Mz1407), which drives expression in all NBs, but not in the NECs of the OPC or LPCs (Wang et al. 2011) (Fig. 1b; Supplementary Fig. 4a and b); Elav-Gal4, which drives expression in all neurons (Supplementary Fig. 4c); and Repo-Gal4, which drives expression in glial cells (Supplementary Fig. 4d).

We found that GFP::Asp^MF expression with Insc-Gal4 could fully restore the size of the entire brain to WT levels, whereas only a marginal rescue was observed for the Elav-Gal4 driver compared to genotyped-match controls (∼14% reduction for Elav-Gal4 compared to 22% reduction for WT vs asp mutants) (Fig. 2i; Supplementary Fig. 2k). No rescue was observed for the glia cell Repo-Gal4 driver (Fig. 2i; Supplementary Fig. 2k). Interestingly, when comparing optic lobe sizes alone, we found that Insc-Gal4 could not fully restore optic lobe size, as these animals still displayed a ∼15% reduction (compared to the 39% reduction observed in WT vs asp mutants). Neither Elav-Gal4 nor Repo-Gal4 showed a significant rescue of optic lobe size either, with a 25–35% reduction in optic lobe size observed in these animals (Fig. 2j; Supplementary Fig. 2l).

The ability of Insc-Gal4 to rescue entire brain size but not optic lobe size is due to an increase in central brain size that compensates for the reduced optic lobe size, as evident in the larger overall entire brain size for the genotyped matched control animals (Insc-Gal4 > UAS-Asp^MF; +/+) compared to WT (asp^t25/+) (Supplementary Fig. 2k and l). It is also in agreement with the spatial expression pattern of Insc-Gal4, which is expressed in most, but not all dividing cells of the optic lobe, the most notable of which are the symmetrically dividing NECs (Wang et al. 2011) (Supplementary Fig. 4a and b). We also suspect that the partial rescue observed for the Elav-Gal4 driver is due to additional weak expression in the NBs and GMCs in the central brain (O’Neill and Rusan 2022).

These results, together with the endogenous expression pattern observed for Asp in the neurogenic larval brain (Fig. 1), strongly support the conclusion that Asp is needed in dividing NSC populations but not postmitotic cells to promote proper brain size. It also confirms that Asp^MF can fulfill most of the brain size specification roles performed by the full-length protein in these cell types.

Transcriptional profiling of asp MCPH brains throughout development

To identify genetic and cellular pathways that are disrupted in asp mutant brains and potentially contribute to the reduced number of cells and the small brain phenotype, we next analyzed the transcriptional profile of the CNS from asp heterozygous controls (asp^T25/+), asp mutant (asp^T25/asp^Df), and asp rescue (ubi-GFP::asp^MF/+; asp^T25/asp^Df) animals at the key developmental stages (third instar larva, P7 pupa, and 3- to 5-day-old adults) used for the analysis above. Using this approach, we reasoned that we could identify the major genetic pathways contributing to brain size and the key developmental stage when these defects occur to prevent proper brain growth (Fig. 3a).

Fig. 3.

Identifying transcriptional signatures in a Drosophila model for MCPH5. a) Experimental workflow for the brain RNA-seq analysis across multiple genotypes [heterozygous control (WT), asp mutant, and asp rescue] and developmental stages [L3 wandering larvae, P7 pupae (∼48 h APF), and adults 3- to 5-day posteclosion], created with BioRender.com. MA plots (logCPM vs logF.C.) for the asp mutant vs WT DES analysis from b) larvae, e) pupae, and h) adults. MA plots (logCPM vs logF.C.) for the asp mutant vs asp rescue DES analysis from c) larvae, f) pupae, and i) adults. Venn diagrams showing the percentage of shared and unique DEGs across asp mutant vs WT, asp mutant vs asp rescue, and asp rescue vs WT comparisons in d) larvae, g) pupae, and j) adults.

After total RNA extraction, samples were subjected to Illumina sequencing and the resulting pair-end reads were aligned to the D. melanogaster genome. A total of 881 million pair-end reads were obtained; 663 million reads (78.1%) were mapped to the fly genome after adapter trimming and verifying sequence quality using FastQC. Biological replicates (n = 4) clustered by genotype and developmental stage based on sample-to-sample distance and principal component analysis (PCA), indicating suitable intergroup and intragroup variation between the correct data sets for downstream analysis (Supplementary Fig. 5).

Impairment of neurogenic pathways and transcription factor networks in the asp mutant larval brain

We then performed pairwise DGE analyses for all genotype combinations at each developmental stage. Between ∼700 and 2,000 genes (coding and noncoding) were found to be significantly differentially expressed among the various combinations (logF.C. of ≥0.5/≤−0.5; adjusted P-value of ≤0.05) (Fig. 3d, g, and j; Supplementary File 1). Larval brains showed a roughly equal split between upregulated and downregulated genes across all genotype comparisons, while both pupal and adult brains showed a significant increase in upregulated genes compared to downregulated genes across asp mutant vs WT and asp mutant vs asp rescue comparisons (Fig. 3b, c, e, f, h, and i; Supplementary Fig. 6). This suggests that the CNS transcriptome is dynamically regulated during development and that asp mutations may significantly alter this landscape in a stage-specific manner.

To gain insight into the biological processes affected upon loss of asp, we utilized DAVID functional annotation to identify clusters of enriched GO and KEGG pathway terms across multiple genotype comparisons and developmental stages (Supplementary File 2) (Huang et al. 2009; Sherman et al. 2022). Our initial plan was to focus on the list of DEGs that were identified in both the asp mutant vs WT and asp mutant vs asp rescue analysis (Fig. 3d, g, and j). We reasoned that since brain size is largely restored in the asp rescue strain expressing the GFP::Asp MF (Asp^MF) in the asp mutant background (Fig. 2b) (Schoborg et al. 2019), it should act as a proxy for WT brains and thus the gene expression signature relevant to brain growth control between the 2 lines should be similar. This would allow us to eliminate any genes whose enrichment may have been due to genetic background differences (e.g. additional effects caused by the overexpression of the GFP::Asp MF) and enable identification of the conserved biological processes shared between both comparisons. It would also allow us to mask any effect imparted by the disrupted mitotic spindles present in both animals, since this is not the primary driver of the small brain phenotype (Fig. 2b and h).

However, given the relatively low amount of gene overlap between these 2 genotype comparisons (267, 177, and 446 genes from larva, pupa, and adults, respectively) (Fig. 3d, g, and j; Supplementary File 1), this analysis returned only a handful of statistically significant functional annotation clusters related to transcriptional regulation and transcription factor activity, stress responses, immune system, lipid metabolism, and cuticle development (Supplementary File 2). Therefore, we next applied DAVID functional annotation to the entire set of DEGs from the asp mutant (M) vs WT (W) (1,438 larval, 1,237 pupal, and 1,576 adult genes) and the asp mutant (M) vs asp rescue (R) (670 larval, 947 pupal, and 1,037 adult genes) comparisons (Fig. 3d, g, and j; Supplementary File 2). We then built an enrichment map network based on these functional clusters using EnrichmentMap in order to identify highly overlapping gene sets and networks between [M vs W] and [M vs R] at each developmental stage (Supplementary File 3) (Merico et al. 2010).

Several interesting networks were identified in this analysis. For larva, we noted a strong enrichment for transcriptional processes linked to neurogenesis, Notch signaling, and eye development and morphogenesis in the downregulated gene sets from [M vs W] and [M vs R] (Fig. 4a). Shared genes enriched in the neurogenesis node included many known Notch players, including Notch (N) receptor, the downstream transcriptional targets belonging to the enhancer of split (E(spl)) complex (e.g. E(spl)mβ-HLH and E(spl)m4-BFM), and the negative regulator earmuff (erm). Other genes include the transcription factor amos, which promotes dendritic neuron formation; H6-like homeobox (Hmx), which is involved in specification of neuronal cell types; and off-track (otk), which is associated with noncanonical Wnt signaling (Supplementary File 3).

Fig. 4.

Transcription factor and signaling networks important for neurogenesis and tissue development are downregulated in asp mutant brains. EnrichmentMap visualization of the a) transcription, neurogenesis, Notch signaling, and tissue development network identified from DAVID functional analysis of the asp mutant vs WT [M vs W] and asp mutant vs asp rescue [M vs R] differentially expressed downregulated larval genes. Nodes represent a set of genes; edges (lines) represent gene overlap between each node. b) Relationships between the 44 genes that comprise the “RNA Pol II core promoter proximal region, sequence-specific DNA binding” node as revealed by GeneMANIA. This gene set is highly enriched in spatial and tTFs and Notch signaling targets that control optic lobe neurogenesis events. Line colors: green, genetic interactions; magenta, physical interactions; blue, colocalization; purple, coexpression; and orange, predicted.

A strong enrichment for other well-known neurodevelopmental transcription factors involved in eye, optic lobe, and overall brain development was also observed in the larval comparison (Fig. 4a and b), including a number of spatial and tTFs that are responsible for orchestrating the progression from neurogenesis to neuronal diversity in the fly visual system (Konstantinides et al. 2022). For example, the shared nodes labeled “RNA polymerase II core promoter proximal regional sequence-specific DNA binding” (GO: 0000978) and “sequence-specific DNA binding” (GO: 0043565) contain 44 and 28 genes, respectively. These genes include Dichaete (D), tailless (tll), sloppy paired 1 (slp1), runt (run), eyeless (ey), odd paired (opa), Optix, twin of eyeless (toy), erm, lozenge (loz), sine oculus (so), anterior open (aop), intermediate neuroblast defective (ind), hamlet (ham), glass (gla), and dachshund (dac), among many others, including the E(Spl) genes identified in the neurogenesis and Notch signaling pathway nodes (Fig. 4b; Supplementary File 3). Only 3 of these transcription factors (ato, Hr3, and E(spl)mdelta) were also found in the asp rescue vs WT comparison, suggesting that impaired neurogenesis through global downregulation of key neurodevelopmental spatial and tTFs and signaling pathways (e.g. Notch) may be a defining feature of asp mutant larval brains.

For the pupal and adult stages, only a few enriched downregulated networks were identified, in agreement with the finding that many fewer downregulated genes were available for this analysis (Supplementary Fig. 6). Interestingly, transcription factor and chromatin-related terms were also identified in the adult, including Histone H4 replacement (His4r), polybromo, early boundary activity 2 (Elba), and the notch signaling-related components insensitive (insv) and transcription factor AP-2 (TfAP-2), while pupa had many fewer downregulated networks, primarily involving oxidoreductase activity (Supplementary File 3).

We next identified upregulated gene networks at each developmental stage between [M vs W] and [M vs R]. Glutathione and drug-related metabolic processes were the only shared networks observed in larval stages. However, several shared networks were found in the pupal and adult comparisons, including membrane, actin cytoskeleton, and chitin-based cuticle development networks. We also noted a shared, strong overlap for immune-related terms at these same developmental stages as well (Supplementary File 3).

Inflammatory pathway activation may be a hallmark of asp mutant brains

While the gene network enrichments for the pairwise [M vs W] and [M vs R] comparisons at each developmental stage were useful for identifying stage-specific biological processes that may be contributing to small brain size in asp mutants, we next wanted to identify processes that were conserved across all developmental stages. This would allow us to determine if there is a common theme or hallmark of asp mutant brains, which could then be tested genetically via mutational analysis to determine whether such shared processes are a cause or a consequence of the small brain phenotype. We therefore performed a similar gene network analysis ([M vs W] and [M vs R]) using EnrichmentMap but did not separate these by developmental stage or expression (upregulated and downregulated). Thus, all 6 comparisons using the entire DEG list were used for the analysis.

The results are shown in Fig. 5. Nine networks were identified, including myosin and actin cytoskeleton, neuropeptide, glutathione and cytochrome P450 metabolism, chitin and cuticle development, proteolysis, sugar metabolism, membranes and cell junctions, oxidoreductases, and immune response (Fig. 5a). However, most of these clusters consist of nodes that are only found in a few (≤3) of the pairwise comparisons and are therefore unlikely to serve as a hallmark of asp mutant brains. Thus, we focused on the networks that contained at least 4 of the 6 pairwise comparisons for multiple nodes. This revealed a single cluster—immune response—as a potential hallmark of asp mutant brains throughout development (Fig. 5a).

Fig. 5.

Biological networks conserved across developmental stages and genotype comparisons highlight inflammation as a hallmark of asp mutant brains. a) EnrichmentMap clustering and visualization of 9 networks identified as significantly enriched across larval, pupal, and adult brains from each genotype comparison. Node coloring for each genotype comparison is shown in the legend; nodes with multiple colors indicate the same DAVID functional annotation classification was found across multiple genotype comparisons and developmental stages. Individual node names are not shown for clarity (see Supplementary File 3 for a full list). b) The immune response network visualized as a linear stack of nodes, with the name of each node representing the DAVID functional annotation identifier used to generate it [GO, KEGG pathway or functional annotation (KW)].

The nodes enriched in the immune system cluster are shown in Fig. 5b and consist of GO terms, KEGG pathways, and DAVID functional annotations that were used by EnrichmentMap to identify the immune response as a significantly enriched network. Node names such as “Toll and IMD signaling pathway,” “innate immunity,” “defense response to Gram-positive bacterium,” and “immunity” are especially prevalent across multiple genotype and developmental stage comparisons (Fig. 5b). The “immunity” and “Toll and IMD signaling pathway” gene sets contain the largest number of genes (44 and 38 genes, respectively) and are enriched in 5 out of the 6 pairwise comparisons. These genes include many well-known players in insect immunity, including the upstream receptors belonging to the peptidoglycan recognition protein (PGRP) family, downstream effectors such as pirk and IKKβ, the activating NF-ΚB transcription factors Relish (Rel), dorsal (dl), and Dorsal-related immunity factor (Dif), and the downstream transcriptional targets belonging to the large family of secreted antimicrobial peptides (AMPs) [Cecropins, Diptericins, Attacins, Bomanins (IMs), etc.] (Supplementary File 3).

Further inspection of the immune-related nodes revealed 2 general trends. First, the immune response was found to be upregulated in virtually all of the pairwise comparisons, with the exception of the larval [M vs R] comparison. This had a significant enrichment of downregulated immune genes (∼10 total) belonging primarily to a small subset of AMPs, such as Attacin-A (AttcA) and Attacin-B (AttB). Second, we noted that the immune response appeared to increase as development proceeded, with a stronger and more consistent (e.g. upregulation) trend readily apparent in pupal and adult stages. These data suggest that an inflammatory response may be a hallmark of the asp small brain phenotype.

Identifying enriched networks of coregulated gene modules in asp mutant brains

To complement our functional enrichment analysis, we also utilized a Drosophila-specific gene signature set consisting of 850 modules of coregulated genes identified through ICA (Rusan et al. 2020). The advantage of this analysis is that it uses gene coexpression patterns identified in flies from thousands of high-throughput data sets and therefore may have additional analytical power to uncover subtle expression perturbations and the underlying gene regulatory networks. This is particularly true in the brain, where it has been used to identify transcriptional signatures unique to individual brain cell types (Rusan et al. 2020) and thus can provide deeper insight into fly-specific brain function and physiology (Wang and Wang 2019). We therefore searched for enriched coregulatory signatures that could be biologically relevant for regulating brain size.

The main strength of the ICA module analysis is that all gene logF.C.s from a genotype pair are used when interrogating ICA modules rather than employing traditional cutoff values (based on logF.C. and adjusted P-value) as in DGE-based analyses. This maximizes expression pattern matching sensitivity for low amplitude (but nonetheless biologically relevant) signatures. In other words, small or subtle changes in logF.C. can be identified as significant if multiple genes part of a coexpressed cluster also have subtle changes and the majority move in the same direction (upregulated or downregulated). This is illustrated by the M₃₉₃₊ module (Supplementary Fig. 7b), which consists of over 400 coexpressed genes. Seventy percent of these genes have logF.C.s that do not meet the logF.C. of ≥0.5/≤−0.5 criteria in both the [M vs W] and [M vs R] comparisons, yet the genes are given higher weights in this analysis and their module achieves a significant Z-score that meets the Z ≥ 3/≤−3 cutoff (Z = −4.9 and −4.4 for [M vs W] and [M vs R], respectively) as a result of their collective downregulation. Conversely, highly significant modules (Z-score of ≥10/≤−10) consist of coexpressed genes with larger logF.C.s (logF.C. of ≥2/≤−2). This is illustrated by M₁₄₆₋, the most highly enriched shared module from the larval analysis. Most of these genes have a large positive logF.C., and thus the module is assigned a highly significant positive Z-score (Z = 15.4 and 18.3 for [M vs W] and [M vs R], respectively) (Fig. 6a; Supplementary Fig. 7a).

Fig. 6.

Identifying coregulated gene network signatures in asp mutant brains. Scatterplots showing significantly enriched coexpression modules identified through ICA analysis of the [M vs W] and [M vs R] data sets from a) larval and b) adult brains. Each module is plotted as a point based on its Z-score enrichment from each analysis. Only significant modules with a Z-score of >3/<−3 are shown. Coexpressed modules having at least 1 immune system-related GO term are colored red, with M₁₁₅₊ outlined with a red box. A subset of the more significantly enriched (Z ≥ 10) modules are also labeled in gray font. The Pearson's correlation coefficient (r) is shown in blue font (2-tailed P < 0.0001). c) Plot of logF.C. vs gene for the M₁₁₅₊ module. Z-score enrichment is shown in green font for each comparison. Each dot is colored based on the mutant vs WT (Mut vs WT, blue) and mutant vs rescue (Mut vs Res, red) value. The red dotted line shows the 0.5/−0.5 logF.C. position. Not all gene names are included on the x-axis for clarity. Relevant immune system genes (l(2)34Fc, Sp212, and Tsf1) are highlighted. d) Word cloud enrichment of the GO terms associated with the M₁₁₅₊ coexpressed module, highly enriched for immune-related, development, and signaling words.

The summary of this analysis is shown in Fig. 6 and Supplementary Fig. 7, along with a file displaying the statistical enrichments and a description of the coexpressed genes and GO terms for each ICA module (Supplementary File 4). Using a Z-score cutoff of Z ≥ 3/≤−3, we evaluated both the [M vs W] and [M vs R] profiles individually as well as the intersection to find common coexpression patterns and filter out potential false positives. A total of 192, 98, and 377 modules met these criteria in larval, pupal, and adult stages, respectively. We also found a strong correlation between the significant modules identified in the [M vs W] and [M vs R] analyses and their Z-score magnitude (Pearson's r = 0.66, 0.70, and 0.87, P < 0.0001 for larva, pupa, and adult, respectively) (Fig. 6a and b; Supplementary Fig. 7d), further evidence that the asp rescue animals largely recapitulate the WT genotype. Interestingly, 97% of adult and 100% of larval and pupal modules were assigned positive Z-scores, meaning that the majority of the genes in each module were upregulated in the comparisons and are in general agreement with the overall expression patterns noted in each developmental stage (Fig. 6a and b; Supplementary Figs. 6a and b and 7d).

From an individual GO perspective, we found significant overlap with our EnrichmentMap analysis (Fig. 5a), thus validating both approaches. For example, M₁₄₆₋, M₃₀₂₋, and M₂₀₈₊ were the most highly enriched (Z ≥ 8) shared modules in larva, pupa, and adult stages (Fig. 6a and b; Supplementary Fig. 7a and d) and contain genes involved in glutathione and sugar metabolism, transport, proteolysis, and cuticle development (Supplementary File 4). GO terms related to “neurogenesis” were found in 7 and 24 modules enriched in larval [M vs W] and [M vs R] profiles, respectively, although only one of these neurogenesis modules was shared between both comparisons. This is also in agreement with our EnrichmentMap analysis (Fig. 4).

Also, we again noted a very strong enrichment of modules containing GO terms related to the immune system at each developmental stage (red dots, Fig. 6a and b; Supplementary Fig. 7d). All of these modules, except for adult M₃₉₃₊, were assigned positive Z-scores in agreement with our earlier DGE findings that immune system genes are upregulated in the asp mutant. However, upon closer inspection of the immune genes found in the negatively assigned M₃₉₃₊ cluster (27 total), there is a clear pattern of behavior where positive regulators of the immune response [e.g. darkener of apricot (doa); Kanoh et al. 2015) display positive logF.C.s, and negative regulators [Enhancer of bithorax (E(bx)) and fat facets (faf); Cronin et al. 2009; Kwon et al. 2009] have negative logF.C.s, consistent with an activated immune response in asp mutant adult brains (Supplementary Fig. 7b). Furthermore, when we examined the 23 intersection modules that were conserved across larval, pupal, and adult brains, three (13%) of these were found to have a significant enrichment of coexpressed immune system genes and associated GO terms, with M₂₁₁₊ and M₉₉₊ enriched in many upregulated AMPs of the immune response (Supplementary File 4).

However, most biological outcomes require input from multiple genes, many of which have distinct functions (and therefore GO terms) that collectively contribute to the final output. This is the primary advantage of coexpression analysis, which is illustrated by one of the most highly enriched modules (M₁₁₅₊) shared across larval, pupal, and adult comparisons (Fig. 6c). M₁₁₅₊ consists of a large number of coexpressed genes (>275), whose GO terms are enriched for words related to immunity (note the positive logF.C. values for the immune system genes l(2)34Fc, Sp212, and Tsf1), development, signaling, metabolism, proteolysis, etc., all of which were identified as separate networks in our EnrichmentMap analysis (Figs. 5a and 6c and d). Although the biological significance of this coexpressed cluster requires further investigation, it suggests that not only are multiple genetic inputs with diverse functions needed for asp-dependent brain growth control but also that these processes may be coordinated at the transcriptional level to achieve the intended biological outcome in the CNS.

A second advantage of the coexpression analysis is identifying new patterns of transcription and potentially new roles for otherwise well-characterized genes. For example, the notch-responsive e(spl) genes that were highlighted in our EnrichmentMap analysis from the larval brain (Fig. 4) were found in 12 different modules from the larval [M vs W] and [M vs R] intersection (M₁₆₆₋, M₂₀₂₊, M₁₆₄₋, M₁₄₈₊, M₁₈₆₊, M₁₇₈₊, M₁₁₃₋, M₃₉₇₊, M₃₉₉₊, M₁₅₋, M₁₀₇₊, and M₁₄₀₊). Interestingly, M₁₆₆₋ was the second highest scoring module for the larva intersection (Z ≥ 8 for both) and contains only the notch-responsive E(spl)mBeta-HLH gene but no other e(spl) member (Fig. 6a; Supplementary File 4). The other 217 genes in this module are involved in a diverse range of biological processes, including cell adhesion, actin cytoskeleton organization, immune system, ribonucleotide and nucleotide metabolism, and cell fate determination. However, 33% of the e(spl)-containing modules have no enriched GO-associated terms, because most of the genes present have no defined function (CG designation on FlyBase). The best example of this is the e(spl)mγ-HLH-containing M₁₁₃₋ module, which contains 125 genes in total, 70 of which are CG genes (Supplementary Fig. 7c and File 4). This module is also interesting because it has positive Z-scores for both [M vs W] and [M vs R], driven by the majority of the CG genes being upregulated in each comparison, yet e(spl)mγ-HLH is significantly downregulated (logF.C.s of −1 and −1.5 in [M vs W] and [M vs R]) (Supplementary Fig. 7c). Together, these results suggest that Notch signaling inputs may be coordinated with many other cellular pathways in parallel, whose identity and biological consequence remain open for investigation.

Inflammation is detectable in asp brains but does not contribute to the tissue disorganization phenotype

Although our analyses highlighted multiple cellular pathways that are disrupted in asp mutant brains, the consistent identification of a transcriptional CNS immune response across development stages using multiple bioinformatics methods led us to further investigate inflammation's role in asp's brain phenotype. The canonical immune response in flies is controlled through 2 pathways, Toll and IMD, that mediate responses to various environmental triggers such as pathogen-associated molecular patterns (PAMPs) and damage-associated molecular patterns (DAMPs) (Supplementary Fig. 8a) (De Gregorio et al. 2002). Downstream activation of the AMPs and other effectors occurs through a family of NF-ΚB transcription factors known as relish (rel), dl, and Dif that mediate IMD (Rel) and Toll (dl and Dif) pathway activation (Lemaitre and Hoffmann 2007).

Previous studies have linked immune system activation to neurodegenerative phenotypes seen in fly models of human neurodegeneration, such as ataxia-telangiectasia (A-T) (Petersen et al. 2012, 2013). These phenotypes consist of vacuole-like structures (holes) in the brain that form as a result of hyperlocalized neuron and glial cell death and loss of the accompanying neuropil (Mutsuddi and Nambu 1998). Hyperactivation of the immune response through AMP production has been shown to induce death of dopaminergic neurons upon loss of Cdk5, and forced overexpression of AMPs independent of Toll or IMD pathway activation can also lead to extensive vacuole formation (Cao et al. 2013; Shukla et al. 2019). Interestingly, we also observed vacuolar structures, prominently disrupted neuropil boundaries, and missing neuronal cell bodies in the optic lobe from asp mutant pupal and adult brains (Fig. 7c and d; Supplementary Fig. 8d and e). This led us to further test whether these phenotypes might be a direct result of immune system activation and, if so, determine their contribution to the asp small brain phenotype.

Fig. 7.

Immune system activation in asp mutant brains leads to upregulated AMP expression. a) qPCR analysis of a subset of AMPs identified as differentially expressed in the RNA-seq analysis from larval brains. Data are shown as the RER; red dotted line shows the RER value for heterozygous control (WT), normalized to 1. b) qPCR analysis in adults from the asp mutant, dif¹; asp^mut, and rel^E38, asp^mut genotypes showing suppression of AMP expression upon loss of NF-ΚB activity. Confocal imaging of adult brains labeled with α-brp (nc82) to visualize the neuropil in c) heterozygous control (WT) and d) asp mutants. Red arrowheads point out disrupted neuropil boundaries; yellow arrowheads show vacuole-like hole structures. The yellow dotted box indicates the adult optic lobe, which is shown in the insets from e) heterozygous control (WT), f) asp mutant, g) rel^E38, asp double mutants, h) dif¹; asp double mutants. Medulla (Me) and lobula (Lo) neuropil regions are labeled. Welch's t-test. ns, P > 0.05; **P ≤ 0.01; ***P ≤ 0.001; ****P ≤ 0.0001. Error bars represent standard deviation. Scale bars = 100 μm c, d); 50 μm e, f, g, h).

We first verified that a subset of IMD and Toll pathway components were significantly upregulated in asp mutants using qPCR, in agreement with our RNA-seq results (∼70% validation rate). Upstream components (IMD, toll, rel, and dif) were mildly upregulated at different developmental stages (Supplementary Fig. 8c). However, we observed significant upregulation of effector AMPs during both larval and adult stages (Fig. 7a and b). IMD-regulated AMPs (AttC, DptB, and CecA2) were consistently upregulated during both stages, with CecA2 and DptB showing their largest upregulation during the larval period. Toll-regulated AMPs (Mtk, IM1, and IM23) showed the strongest upregulation during the adult stage, with IM23 expression increasing nearly 100× from larval to adult brains. AMP activation was also not observed in asp rescue animals, suggesting that this response was specific to asp mutants and not the result of Asp transgene overexpression (Fig. 7a). Interestingly, we did not observe significant expression for these same AMPs during the pupal stage (Supplementary Fig. 8b), suggesting that the AMP response is dynamically regulated throughout development in response to asp loss.

To test whether inflammation is a cause or consequence of the brain phenotypes in asp mutants, we next generated double and triple mutant combinations of asp and the NF-ΚB factors rel (rel^E38) and dif (dif¹). These NF-ΚB mutations suppressed AMP activation in a target-specific manner (Fig. 7b), further confirming the immune response seen in asp mutants. We first focused on the asp adult tissue morphology defects (abnormal neuropil projections, mispositioned medulla, missing cell bodies, and vacuole formation) (Fig. 7c and d), hypothesizing that if these defects were dependent on upregulated AMPs, then their suppression in the NF-ΚB mutants should result in at least partial restoration of WT brain tissue. Qualitative evaluation of adult rel, asp and dif, asp double mutants compared to the single asp mutant did not reveal any significant tissue differences, with all mutant genotypes showing significant optic lobe neuropil disorganization and vacuole-like structures (Fig. 7e–h). Triple mutants (dif; rel, asp) also did not suppress the morphology defects of the asp single mutant (Supplementary Fig. 8f and g). These results show that the tissue morphology defects observed in asp MCPH are not caused by AMP upregulation or other NF-ΚB-dependent gene regulatory networks.

Genetic suppression of inflammation partially rescues the asp brain size phenotype

We next considered whether inflammation might contribute to the asp mutant brain size phenotype. Analysis of rel, asp and dif, asp double mutant combinations revealed a partial rescue of both entire brain and optic lobe volume compared directly to the asp single mutant alone (Fig. 8a and b; Supplementary Fig. 9a and b), with rel, asp and dif; asp double mutants showing a 28 and 26% increase in optic lobe volume, respectively (Fig. 8b; Supplementary Fig. 9b). This partial size rescue becomes even more apparent when factoring in the genetic background influences from the rel and dif single mutants. For example, only a 19 and 34% decrease in optic lobe volume was observed in the rel, asp and dif; asp double mutants compared to the rel and dif single mutants, respectively (note the 42% reduction seen in the asp single mutant compared to WT; Supplementary Fig. 2f). The dif; rel, asp triple mutants showed a similar trend to each double mutant genotype, with a 29% increase in volume compared to the asp mutant and only a 27% reduction compared to the dif mutant (Fig. 8a and b; Supplementary Fig. 9a and b).

Fig. 8.

The NF-ΚB immunity factors relish and dif contribute to the asp mutant brain phenotypes but not through apoptosis. μ-CT volume measurements of a) entire brain and b) optic lobe from heterozygous control (WT), asp mutant, asp, rel double mutant, asp, dif double mutant, and asp, rel, dif triple mutants. c) qPCR analysis of DptB, Mtk, and IM23 AMP mRNA levels following depletion of Dif in either neurons (Elav-Gal4) or glia (Repo-Gal4) in asp mutants. d) Number of apoptotic cells based on DCP-1 staining during the larval stage, normalized for optic lobe volume. e) Adult brains stained with nc82 (α-brp) to visualize the neuropil of the optic lobe from asp mutants expressing the anti-apoptotic baculovirus UAS-P35 protein using the Insc-Gal4 (NBs) and Elav-Gal4 (neurons) driver. Note the severe morphological defects (red arrowheads) present, identical to those seen in the single asp mutants (Fig. 7f). Entire brain f) and optic lobe g) volume measurements (T-ratio) from heterozygous control (WT), asp mutant, and asp mutants expressing UAS-P35 using Insc-Gal4, Elav-Gal4, and Repo-Gal4 drivers. For all size measurement graphs, data are represented as the T-ratio, which normalizes optic lobe volume to thorax width (body size) (Schoborg et al. 2019). asp heterozygous control (asp^T25/+) was set to 1, and the subsequent genotypes were normalized accordingly. n ≥ 5 brains, Welch's t-test. ns, P > 0.05; *P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001; ****P ≤ 0.0001. Error bars represent standard deviation. Scale bars: 50 μm.

To better understand the partial suppression of the asp small brain phenotype upon immune system inactivation, we focused on the mechanism of immune system activation and its potential consequences. Glial cells are the primary modulator of the CNS immune response in flies and vertebrates, although neurons have also been shown to mediate inflammatory responses in fly neurodegenerative models (Cao et al. 2013; Hartenstein and Giangrande 2018; Maitra et al. 2019). We used Repo-Gal4 and Elav-Gal4 to silence Dif expression specifically in asp mutant glia and neurons (>95% knockdown; Supplementary Fig. 9c) and then measured DptB, Mtk, and IM23 AMP levels in adult brains. We found that both glial cells and neurons contribute to the AMP upregulation response in asp mutants, with glial cells contributing a slightly stronger AMP response (Fig. 8c). These data suggest that both postmitotic cell types in the brain are involved in the immune response seen in asp mutant animals.

Finally, we evaluated apoptosis, which can occur as a result of AMP activation in neurons and glial cells (Petersen et al. 2012, 2013; Badinloo et al. 2018; Shukla et al. 2019). Cleaved DCP-1 staining did not reveal a significant difference in the total number of foci between asp mutant and WT larval brains, but we did find a ∼29% increase when normalizing to overall optic lobe volume, although this difference was not statistically significant (Fig. 8d; Supplementary Fig. 9d).

Furthermore, genetically inhibiting apoptosis in NBs (Insc-Gal4), neurons (Elav-Gal4), or glia (Repo-Gal4) in asp mutants using the baculovirus anti-apoptotic protein P35 did not rescue either the brain size or morphology defects (Fig. 8e–g; Supplementary Fig. 9e and f). In fact, we detected a further significant 15 and 18% decrease in the entire brain and optic lobe size in Elav-Gal4/UAS-P35 asp mutant animals compared to the asp mutant alone (Fig. 8f and g). A small but statistically significant increase in brain size was detected in Repo-Gal4 > UAS-P35; asp mutant animals compared to asp mutants alone, although this was likely due to genetic background differences as the Repo-Gal4 genotype matched controls were significantly larger than the Asp heterozygotes (asp^T25/+) used as the WT control (Supplementary Fig. 9e and f). When comparing the percent decrease in the entire brain and optic lobe size between the genotyped matched pairs, we found virtually no difference (49% vs 47% for entire brain and 34% vs 29% for optic lobes; Supplementary Fig. 9e and f). These results suggest that apoptosis in the 3 primary brain cell types (NBs, neurons, and glia) is not the primary driver of the asp mutant brain size phenotype. Thus, the downstream consequences of immune system activation in asp mutants that influence brain size remain unknown.

Discussion

Our primary motivation in this work was to examine the neurodevelopmental time course of the Drosophila asp mutant brain phenotype and identify the genetic and cellular factors contributing to it. Our neuronal cell counts coupled with accurate volume measurements of the CNS suggest that reduced brain size is already evident during the late neurogenic time window of development, which persists during later stages of neural development without significant changes to brain size and neuronal cell number. This agrees with the pattern of Asp expression that we found, which suggests that nearly all neurogenic cell types in the brain require high amounts of asp expression, which then decreases significantly in later stages once the bulk of neurogenesis has been completed.

The key question is ultimately the fate of these missing cells—why are they not being made correctly? Works in vertebrates have identified defects in mitosis, cell cycle length control, centriole duplication, maintenance of apical complex proteins, and premature delamination of radial glial cells as contributors to ASPM MCPH (Fish et al. 2006; Capecchi and Pozner 2015; Jayaraman et al. 2016; Johnson et al. 2018), suggesting that the factors shaping brain tissue size and architecture are complex and involve multiple genetic and cellular pathways operating in parallel.

The advantage of our transcriptome analysis is that it leverages the benefits of omic-based approaches (Wang and Wang 2019) and highlights many potential new pathways involved in brain growth control, thus broadening the scope of the cellular basis of the MCPH. However, untangling these multiple inputs is complicated in human MCPH patients, where access to viable tissue, especially during the neurogenic window of neurodevelopment, can be difficult to acquire (Evrony et al. 2017). One advantage of using the fly system is the ease at which living mutant tissue can be acquired at defined neurodevelopmental stages, and the availability of genetic tools to functionally assess candidate hits to identify the actual pathways and their relative contributions to brain growth and development.

The data presented here provide insights into this approach, both in terms of its advantages and its limitations. To our knowledge, this is the only bulk transcriptomic data set available for asp/Aspm MCPH. A previous study used single-cell RNA-seq (scRNA-seq) to examine the transcriptional profile of an ASPM^−/− ferret model, although the data were used to determine relative proportions of neural progenitor cells rather than examine transcriptional differences between WT and mutant animals (Johnson et al. 2018). Thus, whether the transcriptional signatures we have identified here are conserved in vertebrates remains to be seen, although it provides a suitable data set for comparison. However, one limitation is the lack of cellular resolution in our data set. For example, even the neurogenic larval brain contains a significant amount of postmitotic neurons and glial cells among its many NSC types, limiting the ability to reveal cell-type–specific transcription signatures (Suzuki et al. 2013; Konstantinides et al. 2022). Such data would have allowed for deeper investigation into the cellular basis of the neurogenesis transcription factor network and immune system pathways, for example, and also help identify other contributing pathways that may have been filtered out in the bulk analysis.

Our coexpression-based module analysis was employed to partially overcome this limitation and provide a more thorough evaluation of the gene regulatory networks operating in asp mutant brains. We were encouraged by the fact that from a single GO-level perspective, the biological pathways identified in our function analysis (immune system, neurogenesis, proteolysis, stress and metabolic pathways, etc.) were also found in the module analysis. Also, the module analysis provided greater insight into how these disparate pathways might be coupled at the transcriptional level and integrated into multiple inputs operating in parallel. However, it is important to note that the construction of these coexpressed networks is dependent on the input data sets used (microarray in this case) (Rusan et al. 2020), and further refinement of them could be achieved through additional data sets, especially from the tissues of interest. Thus, there is a possibility that the modules containing a large number of genes (>200) and different GO associations could be further resolved into their own unique modules. Nonetheless, even though this would suggest an uncoupling at the expression level, a high enrichment of the individual modules as a whole would still support the conclusion of multiple biological pathways operating in parallel in asp mutant brains.

Another insight involves the genetic characterization. Our double and triple mutant analysis to evaluate the contribution of the immune system activation we consistently observed across neurodevelopment revealed a partial rescue of the brain size phenotype. We chose to focus on inflammation for reasons outlined above, including its involvement in other fly neurodegenerative phenotypes, and work by Lemaitre and others has elucidated the molecular underpinnings of this response through mutational analysis in Drosophila and is thus fairly well characterized with a number of mutant alleles available (Hedengren et al. 1999; De Gregorio et al. 2002; Lemaitre and Hoffmann 2007; Petersen et al. 2012, 2013; Cao et al. 2013; Kounatidis et al. 2017). Although our data suggest that inflammation can contribute to the asp brain size phenotype, when and how this occurs is currently not known. We did not perturb immune function in a developmental stage manner; thus, it remains an open question as to whether its involvement is more critical during neurogenesis or the later time points. As for the how, factors such as cell number, size, shape, and packing density can affect overall brain volume in vertebrates (Herculano-Houzel et al. 2006, 2007), although the extent to which inflammation can alter these properties is less defined. We tested apoptosis, a common outcome of immune system activation in the fly CNS (Petersen et al. 2012, 2013; Badinloo et al. 2018; Shukla et al. 2019) and a logical pathway for the reduction in total cell number and brain size that we observed in asp mutants, but found it surprisingly does not contribute to the brain phenotypes either. It also remains an open question as to whether the NF-ΚB factors have other transcriptional targets unrelated to the immune response that could be responsible. This is especially true for dif, which has additional developmental patterning roles during embryogenesis (Stein et al. 1998) and whose brains are ∼10% larger than WT.

Another observation from this analysis is that the 2 prominent asp mutant brain phenotypes, tissue morphology (neuropil disorganization) and size, appear to be independent of each other. We currently do not know what is responsible for these morphology defects, although it may be a consequence of the disrupted larval neuroepithelium seen in asp mutant larva animals thought to be a result of disrupted actin cytoskeleton function (Rujano et al. 2013). Interestingly, we did find an enrichment of actin, myosin, and other muscle-related factors in our network analysis for larval and adult stages, suggesting an underlying transcriptional effect on the actin cytoskeleton might also contribute to 1 or both asp mutant brain phenotypes, although we have yet to test this directly. Nonetheless, our data suggest that tissue morphology and size can be uncoupled from one another, providing additional evidence that the biology contributing to these processes is complex and that multiple pathways affected by asp loss collectively contribute to brain growth and development.

One unexpected finding, but one that we speculate is the primary driver of the neurogenic defects and subsequent asp mutant brain phenotypes, was the global downregulation of a number of transcription factors that regulate key neurodevelopmental events, particularly neurogenesis and eye development. A number of studies from the Desplan Lab and others have revealed an intricate transcription factor cascade consisting of spatially and temporally acting factors that ultimately control the transition from progenitor cells to a diverse suite of neurons in the optic lobe (Suzuki et al. 2013; Rossi et al. 2017; Konstantinides et al. 2018, 2022; Kurmangaliyev et al. 2020; Özel et al. 2021). Our larval analysis highlighted many of these factors, including ey, D, tll, opa, slp1, Erm, run, Optix, and toy (Konstantinides et al. 2022). We also found other transcription factors known to play important roles in eye, optic lobe, and lamina development including so, rough (ro), gl, pebbled (peb), spalt major (salm), aop, and dac (Noveen et al. 2000; Adachi et al. 2003; Anderson et al. 2006; Furukubo-Tokunaga et al. 2009; Suzuki et al. 2013; Rossi et al. 2017). Our asp mutants display retina, lamina, and optic lobe defects, the latter of which is the most significant contributor to the brain size phenotype (Schoborg et al. 2019), and are consistent with mutant phenotypes reported for a number of these factors. Eyeless mutants, for example, show optic lobe morphological defects that are strikingly similar to those seen in asp mutants, including reduced medulla, lobula and lobula plate size and orientation, plus defects at the neuropil boundaries, and ectopic fiber bundles (Callaerts et al. 2001). Further work is needed to determine the relative contribution of these spatial and tTFs in the asp mutant brain phenotype and are being prioritized for follow-up.

In addition to these transcription factors, we also found an enrichment of Notch signaling–related components that were also downregulated, particularly the e(spl) complex of transcriptional repressors (Jennings et al. 1994). Notch signaling's role in neurogenesis is well established, operating at multiple points along the temporal progression from NBs to differentiated neurons to determine cell fate in both flies and vertebrates (Campos-Ortega 1993; Chitnis et al. 1995; Shimojo et al. 2008; Paridaen and Huttner 2014). In the fly optic lobe, loss of function Notch and Delta mutants have a smaller medulla and lamina as a result of premature differentiation of NECs into mNBs, with gain of function mutations showing an expansion of the neuroepithelial pool at the expense of NBs (Wang et al. 2011). Notch signaling has also been shown to directly regulate the expression of the tTF slp1 in the medulla (Ray and Li 2022), which was identified in our analysis. Notch also has a role in the terminal differentiation of neurons in the optic lobe of flies, where it cooperates with the tTFs mentioned above (e.g. D, toy, and run) in a birth-order–dependent fashion to generate neuronal diversity in the medulla (Wang et al. 2011; Pinto-Teixeira et al. 2018; Konstantinides et al. 2022). Thus, disruption of Notch signaling in the brain could have a diverse range of cellular consequences that span the entire neurogenesis program of the developing optic lobe. Experiments are currently underway to address these questions, although it is also worth mentioning that our asp mutants show many of the classic Notch mutant phenotypes, particularly in the wing and bristles (Mummery-Widmer et al. 2009). However, our module analysis also found many of the e(spl) genes coexpressed with genes of both known and unknown functions, again suggesting that if Notch signaling contributes to the asp mutant brain phenotype, it likely does so with multiple other inputs as well.

Lastly, the question of how mutations in asp could lead to a downregulation of a number of spatial and temporal optic lobe transcription factors and their regulatory networks remains unknown. One possibility is that Asp may have a direct role in regulating these genes and their networks by acting as a transcription or chromatin factor. Interestingly, Asp and its human ortholog (ASPM) have been shown to localize to the interphase nucleus of Drosophila meiotic I cells and human-cultured cells (Wakefield et al. 2001; Zhong et al. 2005; Higgins et al. 2010), although whether it has a nuclear function (such as transcriptional regulation) remains to be explored. Given that the regulatory hierarchy of these temporal factors is complex and not fully understood (Konstantinides et al. 2022), much more work will be needed to explore this. Another possibility is that Asp regulates the signaling pathways (e.g. Notch) that dictate proper spatial and temporal expression patterns of these factors through direct protein–protein interactions of pathway regulators or perhaps even direct interaction with the tTFs themselves. More work will be needed to test these hypotheses and determine the relative contribution of these regulatory pathways and others in brain tissue architecture and size regulation in flies, which could provide insight into the etiology of human MCPH5.

Data availability

All raw and processed sequencing data generated in this study can be found at the NCBI Gene Expression Omnibus (GEO; https://www.ncbi.nlm.nih.gov/geo/), under accession number GSE244463. Supplemental material is available at figshare (https://doi.org/10.25386/genetics.24058836).

Supplemental material available at GENETICS online.

Funding

This work was supported by grants from the National Heart, Lung, and Blood Institute of the National Institutes of Health (1K22HL137902-01) and an Institutional Development Award (IDeA) from the National Institute of General Medical Sciences of the National Institutes of Health under grant #2P20GM103432. SC and SF are supported by the USDA National Institute of Food and Agriculture , Hatch project #1012152. MBC was supported by the Wyoming Research Scholars Program. Any opinions, findings, conclusions, or recommendations expressed in this publication are those of the author(s) and do not necessarily reflect the view of the NIH, National Institute of Food and Agriculture (NIFA), or the United States Department of Agriculture (USDA).