The Drosophila histone demethylase KDM5 is required during early neurodevelopment for proper mushroom body formation and cognitive function

Mutations in the lysine demethylase 5 (KDM5) family of transcriptional regulators are associated with intellectual disability, yet little is known regarding the spatiotemporal requirements or neurodevelopmental contributions of KDM5 proteins. Utilizing the mushroom body (MB), a major learning and memory center within the Drosophila brain, we demonstrate that KDM5 is specifically required within ganglion mother cells and immature neurons for proper neurodevelopment and cognitive function. Within this cellular subpopulation, we identify a core network of KDM5-regulated genes that are critical modulators of neurodevelopment. Significantly, we find that a majority of these genes are direct targets of Prospero (Pros), a transcription factor with well-established roles in neurodevelopment in other neuronal contexts. We demonstrate that Pros is essential for MB development and functions downstream of KDM5 to regulate MB morphology. We therefore provide evidence for a KDM5-Pros axis that orchestrates a transcriptional program critical for proper axonal development and cognitive function.


INTRODUCTION
Intellectual disability (ID) is reported to affect 1.5% to 3% of the global population and represents a class of neurodevelopmental disorders characterized by cognitive impairments that result in lifelong educational, social, and financial consequences for patients and their caregivers (van Bokhoven, 2011;Leonard and Wen, 2002). ID disorders are diagnosed during early childhood and are defined by an IQ score of less than 70 with deficits in adaptive behaviors (Ropers, 2010). However, despite these profound burdens, little is known regarding the pathogenesis of ID disorders, particularly how disruptions in genetic and neuronal regulatory programs contribute to cognitive and behavioral dysfunction.  (Witteveen et al., 2016). Additionally, a subset of ID-associated KDM5C missense mutations have been shown in vitro not to affect H3K4me3 demethylase activity, yet alter transcriptional outputs (Brookes et al., 2015;Vallianatos et al., 2018). Collectively, these data provide strong evidence that disruption of KDM5 protein function may impact multiple transcriptional pathways critical to neuronal development and function.
Here, we utilize Drosophila, which encodes a single, highly conserved KDM5 ortholog, known as kdm5 or little imaginal discs (lid), to investigate the genetic, neuromorphological, and behavioral consequences of KDM5 loss during neurodevelopment. Our analyses focus on a group of neurons known as Kenyon cells, which form a bilateral, neuropil-rich structure known as the mushroom body (MB). The MB is essential for orchestrating a diverse repertoire of cognitive processes and is thus routinely used to study neuroanatomical changes associated with mutations in ID-related genes ( We show here that kdm5 knockout and shRNA-mediated depletion of kdm5 within GMCs and immature neurons both result in profound defects to MB axonal guidance and growth. Cellspecific depletion of kdm5 additionally results in significant behavioral deficits as revealed by altered avoidance response to Drosophila stress odorant (dSO). Furthermore, using an in vivo transcriptomics-based approach, we identify a subset of genes implicated in GMC fate determination and axon guidance that are downregulated by kdm5 loss within GMCs and immature Kenyon cells. One such gene, prospero (pros), encodes a homeodomain-containing transcription factor with a critical role in promoting proper axon pathfinding and growth. While Pros plays well-established roles in other neuronal cell types, its importance in regulating MB development remains unexplored. We find here that MB-specific knockdown of pros leads to aberrant axonal growth and guidance. We additionally demonstrate that half of KDM5 regulated genes within GMCs and immature neurons are bound by Pros and that kdm5 and pros lie within the same genetic pathway. Our studies thus provide the first in vivo analysis of KDM5 within a specific cell population, revealing a key genetic interaction between kdm5 and pros critical for neurodevelopment.

KDM5 is expressed throughout the CNS and is essential for proper MB morphology
To assess the neurodevelopmental consequences resulting from KDM5 loss, we examined MB morphology of animals that were homozygous for a kdm5 null allele, kdm5 140 (Drelon et al., , 2019. As homozygous kdm5 140 animals fail to eclose from their pupal cases, we performed our immunohistochemical analysis on pharate adults, which externally appear indistinguishable from wild-type animals . Following a well-established classification scheme used by others (Gombos et al., 2015;Kelly et al., 2016;Michel, 2004), MB defects were categorized as impacting MB growth and/or guidance, with the former defined by a stunted, overextended, or absent lobe and the latter by a misprojected lobe. Staining using an antibody specific to the NCAM-like cell adhesion molecule Fasciclin 2 (Fas2) revealed highly penetrant MB abnormalities, with ~70% of animals exhibiting growth and guidance defects of the α/β lobes (Fig. 1A,B). Because animals specifically lacking KDM5 demethylase activity have phenotypically normal α/β lobes (Zamurrad et al., 2018), these data demonstrate that KDM5 is required for maintaining MB morphology independent of its canonical enzymatic function.
To assist in our functional understanding of KDM5, we sought to investigate its expression throughout CNS development. To facilitate these analyses, we used CRISPR/Cas9 to generate a strain containing a 3xHA-tag fused to the endogenous locus of kdm5 and demonstrated by western blot that the KDM5:HA protein is expressed at the same level as endogenous KDM5 from wild-type animals (Fig. 1C). We also confirmed that KDM5:HA was expressed in adult MB Kenyon cells by co-expression of a nuclear membrane-localized GFP reporter, UNC-84:GFP, with the MB driver OK107-Gal4 (Fig. 1D). Immunostaining of the central nervous system (CNS) of wandering third instar larvae revealed that KDM5 was localized to cortical nuclei while absent from neuropilrich regions marked by the ubiquitous presynaptic active zone marker Bruchpilot (Brp) ( Fig. 2A).
KDM5:HA localized to cortical nuclei across a variety of cell types, including neurons ( Fig. 2B), neuroblasts (NBs) and presumptive GMCs (Fig. 2C). We also examined KDM5:HA expression in the adult brain, where KDM5:HA appeared to be similarly localized to cortical nuclei while absent from neuropil-rich regions, such as the antennal lobes and both the dorsolateral and ventrolateral protocerebra (Fig. 2D,E). Given the broad expression pattern of KDM5:HA within a variety of cell types, KDM5 may regulate a range of neural processes, from NPC division and axonal growth to neuronal maturation and function.

KDM5 is required within neural precursors and immature neurons for proper MB morphology and behavior
To define the functional requirements of KDM5 during MB development, we utilized an inducible kdm5 shRNA transgene that we and others have shown effectively reduces KDM5 levels broadly within all MBNBs, MB-GMCs, and Kenyon cells throughout development using OK107-Gal4 (Fig. 3A,B). This resulted in significant neuromorphological defects of the α/β MB lobes, with the major phenotype being an overextension of the β lobes across the midline (Fig. 3C,D). We next assessed the consequences of kdm5 depletion within distinct and overlapping subsets of mature Kenyon cells. We hypothesized that if KDM5 functioned within mature, post-mitotic Kenyon cells to maintain MB morphology, then its depletion would lead to neuronal abnormalities.
Given that concomitant depletion of KDM5 within MBNBs, MB-GMCs and Kenyon cells, but not within mature Kenyon cells, resulted in α/β lobe abnormalities, we suspected that KDM5 may function within NPCs to guide MB development. Knockdown of kdm5 using two independent NBrestricted Gal4 driver lines, worniu-Gal4 (wor-Gal4) and inscuteable-Gal4 (insc-Gal4), resulted in profound α/β lobe defects (Fig. 3C,D). Notably, a significant proportion of these brains simultaneously displayed both growth and guidance defects of the α/β lobes. Although wor-Gal4 expression is NB specific, KDM5 depletion and GFP expression continued to be observed in presumptive GMCs and post-mitotic cells surrounding each NB (Fig. 3E). This is likely attributed to perdurance of the Gal4 activator protein and/or the kdm5 shRNA. KDM5 could therefore be required within the NB, GMC or even post-mitotically within the immature neuron for proper α/β Kenyon cell neurodevelopment.
To determine if KDM5 is indeed required within NBs for proper MB development, we used This driver did not appear to drive expression within the majority of NBs observed, including within the MBNBs (Fig. 4A). When kdm5 was knocked down using R71C09-Gal4, KDM5 was dramatically depleted within presumptive GMCs and immature neurons (Fig. 4B). We also noted that R71C09-Gal4 strongly drove expression in newly born α/β Kenyon cells of recently eclosed adults whose axons lie within the core fibers of the MB pedunculus, which was consistent with its strong expression in immature neurons (Fig. 4C). Depletion of kdm5 using this driver resulted in profound α/β lobe defects, indicating that KDM5 is functionally required within GMCs and immature MB neurons for proper axonal development (Fig. 4D). As knockdown using R71C09-Gal4 resulted in MB abnormalities, we next used this driver to re-express kdm5 in GMCs and immature neurons of kdm5 140 animals. Re-expression of kdm5 using R71C09-Gal4 significantly rescued the defects we had previously observed in kdm5 140 animals, further demonstrating that KDM5 acts within GMCs and immature neurons to promote proper MB development (Fig. 4E).
To assess whether these MB defects were associated with behavioral deficits, we utilized a robust two-choice behavioral assay which measures avoidance to Drosophila stress odorant (dSO), an innately repulsive sensory cue produced by conspecifics upon mechanical agitation  1988). Importantly, the temperature shifts required to induce the TaDa system in a kdm5 140 animals did not alter the frequency or type of structural MB defects observed (Fig. S1A,B).
Carrying out TaDa from five biological replicates, we identified a total of 636 differentially expressed genes (DEGs) using a statistical cutoff of FDR < 0.05, 438 of which were downregulated and 198 of which were upregulated (Fig. 5B, Table S1). Consistent with previously  Table S2). The finding that ribosomal protein genes, such as RpS2, RpS24, RpS28b, RpL3, Rpl23, RpL39, and RpL41, were altered in our conditions. To ensure that changes to gene expression were reflective of the severe MB phenotypes we had previously observed, we induced the expression of kdm5 shRNA and dam:pol ll for six days, beginning during the early second-instar larval (L2) stage, to allow for sufficient depletion of KDM5 within MB-GMCs and immature α/β Kenyon cells (Fig. 5D). This temporallytargeted knockdown strategy recapitulated the adult α/β lobe phenotypes we had previously observed for constitutive kdm5 knockdown with R71C09-Gal4 (Fig. S4A,B). From three biological replicates and a statistical cutoff of FDR < 0.05, this TaDa experiment identified 1069 DEGs, 659 of which were downregulated and 410 of which were upregulated (Fig. 5E, Table S3). Compared to the kdm5 140 TaDa, the greater number of DEGs from the kdm5 RNAi TaDa could be attributed to the extended duration of the kdm5 knockdown and induction of Dam:Pol II. Additionally, we observed larger changes to gene expression, with an average 5.5-fold decrease among downregulated genes and a 4.3-fold increase among upregulated genes (Fig. 5E, Table S3). This could be ascribed to the acute loss of kdm5 expression due to RNAi-mediated depletion which would decrease the likelihood of compensatory changes occurring.
To obtain a list of high-confidence genes regulated by KDM5, we compared the DEGs  Table S6). These data suggest that Pros could be a key mediator of KDM5 function in GMCs and immature neurons. Our data also implicate Pros in the activation of genes, as the majority of Pros-bound genes were downregulated upon KDM5 depletion (Fig. 7B).
While roles for Pros in regulating NB and GMC asymmetric division are well-described As kdm5 or pros depletion within MB progenitor and Kenyon cells each independently result in significant axonal defects, we next assessed whether kdm5 and pros functioned within the same genetic pathway. To test this, we generated animals heterozygous for null alleles of both kdm5 and pros, thereby simultaneously reducing levels of both proteins. Animals heterozygous for either kdm5 140 or pros 17 did not display significant MB defects. In contrast, adults that were transheterozygous for kdm5 140 and pros 17 displayed MB abnormalities, albeit less severe than those of kdm5 140 homozygous pharates or OK107-Gal4-mediated pros knockdown adults alone (Fig 7E,F). These data are consistent with KDM5 and Pros functioning synergistically to affect MB development. KDM5 is therefore likely to function within GMCs and immature neurons to regulate levels of pros and its downstream targets, thereby driving MB development and promoting proper cognitive function (Fig. 7G).

DISCUSSION
Here, we utilize an in vivo transcriptional approach coupled with neuromorphological and behavioral analyses to demonstrate that KDM5 is essential for the development and function of the adult MB. Although previous work in murine models has shown that knockout of the kdm5 ortholog Kdm5c leads to dendritic spine abnormalities in cortical brain regions, these studies Although we demonstrate a genetic link between kdm5 and pros in regulating MB development, the exact mechanism(s) through which KDM5 accomplishes this remains unknown.
We predict, however, that the regulation of pros by KDM5 is independent of its canonical histone

DECLARATION OF INTERESTS
The authors declare no competing interests.

Lead Contact
Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Julie Secombe (Julie.Secombe@einsteinmed.org).

Materials Availability
The kdm5:HA strain generated in this study is available from the Lead Contact without restriction.
DamID-Seq data have been deposited in the Gene Expression Omnibus (GEO) under accession number GSE156010.

Data and Code Availability
The code supporting the current study is available at http://marshall-lab.org/code/.

Fly Strains and Genetics
A detailed list of the genotypes of the flies used in each figure is included in the Key Resources Table. All GAL4 and GMR GAL4 lines were generated at the Janelia Research Campus/HHMI

Cloning and Transgenesis
To tag the endogenous kdm5 locus with three in-frame HA epitope tags, we used CRISPR/Cas9-mediated knock-in. The HA epitopes and the homology arms, carrying a synonymous mutation for the PAM sequence, were PCR amplified from a clone containing the A similar protocol was followed for immunostaining of pharate adult and 3-to 5-day old adult brain tissue with the following exceptions. Pharate adult heads or whole adult animals were allowed to incubate in fixation buffer (4% paraformaldehyde in 0.2% PBT) at 4 ºC for 3-hrs. Heads or whole animals were then washed in 0.2% PBT for three cycles of 15-min at RT. Brains were then dissected from the fixed tissue and blocked for 30 minutes at RT. Antibody incubation and mounting was identical to that described above.
The following primary antibodies were used: mouse anti-Fas2

Western Blotting
Western analysis was carried out as previously described (Drelon et
All MB confocal stacks were taken under either 2X or 2.5X zoom, in a 1024 x 1024 configuration and using 1 μm resolution. Image stacks were processed with Figi (ImageJ). Figures were composed using Microsoft Powerpoint.

Targeted DamID and Analyses
To profile Pol II occupancy in GMCs and immature neurons of kdm5 140 pupae, Plus Kit. DpnI digestion, PCR adaptor ligation, DpnII digestion, and PCR amplification were performed as described. DNA was sonicated using a Diagenode Bioruptor for 8-10 cycles (5-min at high power, 30-s on/30-s off) and analyzed using an Agilent Bioanalyzer. DamID adaptor removal and DNA cleanup were performed as previously described (Marshall et al., 2016) and samples were submitted to BGI for sequencing.
Sequencing libraries were prepared at BGI Genomics following a ChIP-seq workflow. DNA fragments were first end-repaired and dA-tailed using End Repair and A-Tailing (ERAT) enzyme.
Adaptors were then ligated for sequencing and ligated DNA purified using AMPure beads. DNA was then PCR amplified with BGI primers for 8 cycles and PCR purified with AMPure beads. DNA was then homogenized, circularized, digested, and again purified. DNA was then prepared into proprietary DNA nanoballs (DNB™) for sequencing on a BGISEQ-500 platform with 50 bp read length and 20 M clean reads.

Drosophila dSO Avoidance Assays
Drosophila dSO avoidance assays were carried out as previously described with modifications (Fernandez et al., 2014;Suh et al., 2004). Briefly, 24 hours before the experiment 100 wild-type adult flies ("emitters") as well as 60-65 experimental or control flies ("responders") were each transferred to standard food vials and housed at 25 ºC. For each experimental or control cohort, 100 emitters were transferred to an empty tube ("dSO vial") and vortexed at maximum speed in 15-s bouts, at 5-s intervals, for 1-min and then removed. The dSO vial was immediately cleared and placed in one arm of a T-maze (CelExplorer Labs) and an empty vial in another. Responders were then deposited into the T-maze elevator and allowed to rest for 1-min.
Responders were then given 1-min to choose between the dSO vial and the empty vial. The locations of the dSO and empty vials were alternated between trials to control for potential environmental confounders. Additionally, a separate cohort of emitters was used for each trial.
The Performance Index ("PI") was calculated by subtracting the number of flies within the dSO vial from those within the empty vial and dividing by the total number of flies.

Statistical Analyses
For MB morphological analyses, results are presented as bar plots for which percentage of brains with MB lobe growth and/or guidance defects are calculated. For these analysis, N = number of brains examined. "Growth defects" were defined by an overgrown, stunted, or absent lobe and "guidance defects" were defined by a full or partially misprojected lobe. In the case where both defect types were observed in a single brain, the defect was categorized as a "growth and guidance defect." All MB statistical analysis was performed using GraphPad Prism 8.4 (GraphPad Software, Inc., Ca, US). A 2 x N contingency table was used when comparing MB defects of more than two genotypes, where N = number of genotypes, and significance was determined using a chi-square test with either a Yates' or Bonferroni correction with * = P < 0.05, ** = P < 0.01, *** = P < 0.001, and **** = P < 0.0001.