Traip controls mushroom body size by suppressing mitotic defects

Microcephaly is a developmental growth disorder characterized by reduced cerebral cortex size and neuron number. Mutations in about 40 genes are linked to microcephaly, ranging from microcephaly, primary, hereditary (MCPH; Jayaraman et al., 2018), in which only brain size is reduced, to more severe forms, such as Seckel syndrome (SCKL), in which both brain and body are reduced in size (Khetarpal et al., 2016). Most microcephaly disorders are linked to mutations in genes with one of two functions: centrosome and mitotic spindle functions, or DNA damage repair (DDR). Mutations in centrosome/mitotic spindle genes are thought to disrupt spindle structure and orientation, leading to defects in chromosome segregation or cell cycle progression and, ultimately, neural progenitor cell (NPC) loss through premature differentiation or cell death (reviewed by Nano and Basto, 2017). Mutations in DDR genes are thought to increase DNA damage, activate checkpoint signaling and lead to increased genome instability, cell cycle lengthening and the ultimate loss of NPCs (reviewed by Bianchi et al., 2018). However, most microcephaly genes have not been directly studied in the context of brain development and the cellular etiology of microcephaly has been inferred from studies in non-neuronal contexts. Thus, we aimed to study the microcephaly gene TRAIP in an animal to reveal the etiological mechanisms of disease.

TRAIP encodes a RING domain E3 ubiquitin ligase that localizes to the nucleus and functions in DDR. In humans, mutations in TRAIP cause SCKL with severe microcephaly and reduced body size (Harley et al., 2016). Cultured fibroblasts from individuals with TRAIP mutations have an elongated cell cycle and impaired DNA repair, suggesting that TRAIP mutations cause microcephaly as a result of reduced proliferation, premature differentiation, and apoptosis of NPCs (Harley et al., 2016). A mouse Traip mutant is early embryonic lethal (Park et al., 2007) and Drosophila mutants for the TRAIP ortholog nopo (herein called traip) are maternal-effect lethal (homozygous null mothers lay eggs that arrest in the first few embryonic cell cycles; Merkle et al., 2009), whereas null mutant Caenorhabditis elegans are viable (Sonneville et al., 2019). No studies of TRAIP function in developing brains have been published.

TRAIP has several functions related to DDR (reviewed by Wu et al., 2020). TRAIP localizes to double-strand breaks, where it recruits other factors and promotes H2B monoubiquitylation to stimulate repair (Han et al., 2019; Soo Lee et al., 2016). TRAIP also localizes to stressed replication forks (Feng et al., 2016; Hoffmann et al., 2016) where it removes DNA-protein crosslinks (Larsen et al., 2019) and regulates the choice between NEIL3 or FANC/BRCA-mediated repair pathways (Wu et al., 2019). In addition to these interphase DDR functions, TRAIP also promotes sister chromatid separation during mitosis. A cell entering mitosis may have under-replicated sister chromatids (URSCs), which contain loci that failed to complete replication during interphase, and these URSCs must be resolved to prevent DNA bridges during anaphase. TRAIP initiates the resolution of URSCs by ubiquitylating the MCM7 subunit of CMG helicase at mitotic entry, thus triggering fork unloading and allowing sister chromatid resolution via mitotic DNA repair synthesis (Deng et al., 2019; Priego Moreno et al., 2019; Sonneville et al., 2019; Villa et al., 2021).

However, most previous studies on TRAIP function were performed in cell culture, Xenopus egg extracts, and the early embryos of Drosophila and C. elegans, and thus it is unclear which TRAIP functions are essential for neurogenesis. In this study, we take advantage of the well-characterized brain structure and genetic tractability of Drosophila to establish a model of traip mutant microcephaly. In this context, Traip functions in NSCs during mitosis to suppress DNA bridges and prevent premature NPC loss, pointing to a role in URSC resolution as crucial for proper brain size.

We generated a new full coding sequence deletion of traip using CRISPR (traip^Δ). We use traip⁻ to refer to traip^Δ in combination with either the previously described null allele traip^Exc142 (Merkle et al., 2009) or the deficiency Df(2R)Exel7153 (Parks et al., 2004), as both allelic combinations showed identical brain phenotypes.

To test whether traip controls brain size, we used μ-computed tomography (μ-CT; Schoborg et al., 2019) to compare control and traip⁻ adult brains using normalized volumetric analysis (Fig. 1A). Although optic lobes were not different (Fig. 1B), traip⁻ central brain volume was slightly reduced, and this reduction was rescued by a ubiquitously expressed GFP-tagged traip transgene (ubi-GFP::traip; Fig. 1C). This suggested that traip is required for fully proper brain development.

To determine whether any specific brain subregion was disrupted, we examined neuropil (Cadherin-N; CadN) and axon tracts (Neuroglian; Nrg) in traip⁻ brains. All brain regions appeared qualitatively similar (Fig. S1A), except for a 100% penetrant size and structure defect in traip⁻ mushroom bodies (MBs), a pair of neuropils that mediate higher-order functions in the insect brain (Fig. 1D; Heisenberg, 2003; Modi et al., 2020). In Drosophila, each MB arises from four neuroblasts (MB-NBs), which divide continuously from embryonic development through late pupal stages to produce roughly 2500 Kenyon cells (KCs) per hemisphere (Fig. 1E). These KCs are positioned on the posterior dorsal side of the brain and project axons in a stereotypic, bifurcated pattern to form the γ (first born), α′/β′, and α/β (last born) lobes of the MB (Fig. 1D).

The MBs of traip⁻ adult brains were misshapen and reduced in size (Fig. 1G), with additional axon-related defects: 50% with axon tracts exiting the lobes (0% in controls), 93% with lobes fused at the midline (14% in controls), and 29% missing α or β lobes (0% in controls). traip⁻ α/β lobe volume, identified by FasII (Fas2) staining, was reduced by 70% (Fig. 1H,I), and traip⁻ full MB volume, identified by mCD8::GFP expression driven in KCs by OK107-GAL4, was reduced by 37% (Fig. S1B,C). The more severe size reduction of the last-born, FasII-positive α/β lobes indicates that the traip⁻ MB defect progressively accumulates through development. Cross-sectional area was also reduced in FasII-positive α lobes by 84% (Fig. 1J) and OK107-GAL4>mCD8::GFP α′/α lobes by 54% (Fig. S1D).

MB lobe size reduction was consistent across all traip null mutant combinations (Fig. S1E), whereas hypomorphic mutants (Merkle et al., 2009) had wild-type MB lobe size (Fig. S1E,F). MB lobe size was rescued by ubiquitously expressing GFP::Traip (Fig. S1G,H), and importantly, by specifically expressing GFP::Traip in the MB-NBs and KCs via OK107-GAL4 (Fig. 1I-K). Thus, MB lobe size reduction is linked to cell-autonomous loss of Traip function in the MB-NBs and/or KCs.

Reduced traip⁻ MB lobe size suggests reduced KC number. We used machine learning to count KCs with OK107-GAL4>NLS::mCherry (Fig. 1L). Each traip⁻ hemisphere contained an average of 699 KCs, a 62% reduction compared with controls, which had 1852 KCs (Fig. 1M). KC number in traip⁻ was rescued by OK107-GAL4>UAS-GFP::traip (Fig. 1L,M). Regression analyses showed a significant linear correlation between KC number and both MB volumes (Fig. S2I) and cross-sectional areas (Fig. S2J). Having shown that α lobe cross-sectional area measurement was a suitable proxy for KC number, we used this measurement for the remainder of this study.

To determine whether the traip⁻ MB lobe defects were developmental (primary microcephaly-like) or neurodegenerative (secondary microcephaly-like; Passemard et al., 2013), we investigated pre-adult stages (Fig. 2A). In the third instar larval stage, when only γ lobes are present, traip⁻ MB lobe size was not significantly reduced compared with controls (Fig. 2B). However, during pupal stages there was a significant reduction in MB lobe size of traip⁻ compared with controls (Fig. 2C). Note that the large reduction in MB cross-sectional area between larval and 24 h after pupal formation (APF) stages is due to extensive developmental remodeling of the short, thick larval γ lobes during metamorphosis, which are replaced by longer, thinner α lobes during the early pupal stages (Lee et al., 1999). Thus, traip⁻ exhibits primary microcephaly-like defects, as MB lobe size reduction arises during development.

We investigated the possibility of neurodegeneration by comparing MBs of 3-day-old and 50-day-old adults (Fig. 2D). MB lobe size increased slightly with age in controls (Fig. 2E), reflecting neuropil reorganization in response to life experience (Heisenberg et al., 1995). In contrast, traip⁻ MBs neither grew nor shrunk with age (Fig. 2E). Thus, traip⁻ MB lobe size reduction is not neurodegenerative and does not reflect secondary microcephaly.

We hypothesized that the reduced KC number and MB lobe size observed in traip⁻ could arise in two ways: (1) a defect in the highly mitotic MB-NBs resulting in fewer KCs born, and (2) a defect in the differentiated KCs resulting in cell death. We began to explore which cell types express Traip, using CRISPR to knock in mNeonGreen at the endogenous traip locus (mNG::traip), which resulted in a fly line with wild-type MB morphology (Fig. S2A). mNG::Traip was expressed in all third instar larval central brain neuroblasts (CB-NBs), which include the MB-NBs (Fig. 3A). In addition, mNG::Traip expression persisted in the ganglion mother cells (GMCs) and their immediate daughter neurons (Fig. 3B). However, mNG::Traip was absent from regions of mature neurons in larval brains (Fig. 3A), and was not detectable in adult brain neurons (Fig. S2B). These data suggest that Traip primarily functions in proliferating brain cells.

To determine which cell types require Traip function, we used cell type-specific expression of GFP::Traip in attempt to rescue traip⁻ MB lobe size. First, we used wor-GAL4 to drive UAS-GFP::traip in NBs, which rescued MB lobe size (Fig. 3C,D). However, this did not fully exclude a role for Traip in neurons, as wor-GAL4>GFP::Traip was also found in GMCs and newly born neurons (Fig. S2C). We next tried expressing GFP::Traip in post-mitotic neurons using elav-GAL4, but discovered strong expression in the neuroblasts (NBs) and GMCs (Fig. S2D). Finally, we used nSyb-GAL4 to drive GFP::Traip expression in mature neurons (Fig. S2E), which failed to rescue MB lobe size (Fig. 3E,F). Thus, Traip function in neurons alone is insufficient to prevent microcephaly. We conclude that the primary function for Traip is in proliferating brain cells (NBs and GMCs), although we could not fully rule out the possibility that undetectable levels of Traip also function in post-mitotic neurons.

We next hypothesized that traip⁻ reduction in KC number and MB lobe size is due to premature loss of MB-NBs. The KCs of each MB arise from four MB-NBs that are easily identifiable at 24 h APF, as nearly all other NBs have been lost by this stage (Ito and Hotta, 1992; Truman and Bate, 1988). Controls maintained four MB-NBs throughout pupal development (Fig. 4A,B). In contrast, traip⁻ progressively lost their MB-NBs, with an average 2.2 MB-NBs per hemisphere at 24 h APF decreasing to 0.9 by 72 h APF, and this loss was rescued by OK107-GAL4>UAS-GFP::traip (Fig. 4A,B).

To explore the role of caspase-dependent cell death in MB-NB loss, we used Df(3L)H99, which deletes four pro-apoptotic genes (grim, rpr, hid and skl; Abbott and Lengyel, 1991). MB-NB loss was suppressed in traip⁻ with Df(3L)H99 compared with traip⁻ alone at 24 h APF (3.1 versus 2.2; Fig. 4A,B), although this effect was not observed in later stages. Df(3L)H99 also improved traip⁻ MB lobe size more than 2-fold (Fig. 4D,E) and improved KC number by 1.4-fold (Fig. 4F,G). Because Df(3L)H99 deletes other genes, we used RNAi against the effector caspase Drice, which improved traip⁻ MB lobe size by 1.6-fold (Fig. 4E,H). We also stained control and traip⁻ for cleaved caspase Dcp-1 at the 24 h APF stage, observing one actively dying MB-NB in traip⁻ (Fig. 4I,J). Both controls and traip⁻ had similar numbers of Dcp-1-positive KCs or GMCs per hemisphere (Fig. 4K). However, controlling for the total number of KCs suggested that traip⁻ had a significantly higher proportion of actively dying KCs and/or GMCs compared with controls (Fig. 4L). Together, these data support a model in which Traip prevents premature loss of MB-NBs (and possibly KCs and/or GMCs), and that in the absence of Traip these cells are lost in part via caspase-dependent cell death.

After noticing abnormal nuclei while scoring traip⁻ MB-NB number, we characterized nuclear phenotypes in 24 h APF MB-NBs to gain insight into defects potentially upstream of MB-NB loss. Control MB-NBs had nuclei with relatively smooth, spherical nuclear lamina morphology and no other defects (100%; Fig. 5A). In contrast, some traip⁻ MB-NBs appeared apoptotic, with abnormally condensed DAPI staining (11%; Fig. 5B), and others frequently had nuclear defects, including irregular, crenellated nuclear lamina morphology (12%; Fig. 5C), micronuclei (11%; Fig. 5D), multiple nuclei (9%; Fig. 5E) or extremely large nuclei suggestive of polyploidy (9%; Fig. 5F). These traip⁻ phenotypes were rescued by OK107-GAL4>UAS-GFP::traip (Fig. 5G).

As these nuclear defects could arise during mitosis, we used high-resolution live fluorescence microscopy to analyze mitotic dynamics of third instar larval CB-NBs. We chose to image general CB-NBs because of technical constraints that limited the ability to image MB-NBs properly at high resolution. Unlike control NBs (Fig. 5H, Movie 1), 26% of traip⁻ NBs had prominent chromosome bridges, where sister chromatids appeared attached and did not effectively separate at anaphase onset, but rather became stretched across the midzone through anaphase before eventually separating (Fig. 5I,J, Movies 2,3). Furthermore, traip⁻ larval brains often had cells with abnormally large nuclei (Fig. S3A), reminiscent of the suspected polyploid cells seen in traip⁻ pupal brains (Fig. 5E). We did not observe any other obvious mitotic defects in traip⁻ NBs. Expressing GFP::Traip via wor-GAL4 rescued these mitotic bridge defects (Fig. 5K, Movie 4). We reasoned that traip⁻ mitotic DNA bridges could block cytokinesis, resulting in multiple nuclei or polyploidy, or else lead to chromosome breakage, resulting in micronuclei.

We also used live imaging to explore potential mitotic delays in traip⁻ NBs. Measuring the time from anaphase onset to the moment of complete furrow constriction showed no difference between controls, traip⁻ and wor-GAL4>UAS-GFP::traip rescues, and the presence or absence of DNA bridges in traip⁻ NBs did not affect mitotic timing (Fig. 5L). However, we were unable to identify abscission definitively, and in one case a traip⁻ NB with a DNA bridge experienced a period of disrupted cortical activity (Movie 3), suggesting a possible delay in abscission. We also found no difference in the duration from prophase onset to complete furrow constriction for CB-NBs (Fig. S3B). In fixed larval brains, the mitotic index of control and traip⁻ CB-NBs were similar (Fig. S3C). In 24 h APF brains, the mitotic index of traip⁻ MB-NBs was lower than controls (20% traip⁻ versus 35% controls); however, when MB-NBs with nuclear defects (Fig. 5B-F) were excluded from this analysis, the mitotic index of phenotypically normal traip⁻ MB-NBs was not different from controls (32% traip⁻ versus 35% controls; Fig. S3D). Thus, the reduced mitotic index of traip⁻ MB-NBs could be explained by an inability of polyploid, multinucleate and apoptotic cells to enter mitosis rather than a function for Traip in directly controlling mitotic or cell cycle timing.

We also found evidence suggesting persistent aneuploidy in traip⁻ MB-NBs. Control pupae expressing both mCD8::GFP and NLS::mCherry via OK107-GAL4 had similar fluorescence levels across the field of mature KCs (Fig. 5M). In contrast, traip⁻ pupae often contained one or more clusters of KCs that had either double the signal or no signal for the GFP or mCherry marker (Fig. 5N). These aberrant KC clusters were reminiscent of clonal cell clusters, suggesting that their MB-NB progenitors became aneuploid, either losing or gaining a copy of one transgene, and then continued generating daughter cells. These clones were never observed in controls, whereas traip⁻ brains had an average of 1.3 clones per hemisphere at 24 h and 1.7 clones per hemisphere at 72 h APF (Fig. 5O). The average number of KCs per clone in traip⁻ brains increased from 15 at 24 h to 37 at 72 h APF (Fig. S3E), with some clones containing hundreds of KCs. OK107-GAL4>UAS-GFP::traip suppressed clone formation (Fig. 5L). Reasoning that these clones could arise via either whole-chromosome mis-segregation or through chromosome breakage and partial mis-segregation, we stained for γH2Av to identify double-stranded DNA breaks in 24 h APF MB-NBs. Whereas controls (Fig. S3F) and most traip⁻ (Fig. S3G) MB-NBs had few γH2Av puncta, a small subset of traip⁻ MB-NBs had extremely elevated γH2Av puncta (Fig. S3H,I). These data are consistent with traip⁻ mitotic DNA bridges leading to aneuploid MB-NBs, which often continue producing daughters before eventually being lost.

TRAIP has a known role in resolving URSCs at mitotic onset (Deng et al., 2019; Priego Moreno et al., 2019; Sonneville et al., 2019), and URSCs are predicted to form a special class of mitotic DNA bridges called ultrafine DNA bridges (UFBs; Liu et al., 2014). Human FANCD2 localizes as puncta at the points where UFBs connect with sister chromatids (Naim and Rosselli, 2009), and although FancD2 localization has not been well-characterized in Drosophila, there is precedence for invisible DNA tethers in Drosophila cells (Royou et al., 2010). Thus, to test for UFBs we generated transgenes encoding fluorescently tagged FancD2. We found that mNG::FancD2 puncta localized to mitotic DNA bridges in many traip⁻ NBs (7/12; Fig. 6B), in addition to weakly localizing to mitotic chromosomes in both controls and traip⁻ NBs (Fig. 6A,B). The observed FancD2 on chromosome bridges suggests the presence of UFBs and is consistent with a role for Traip in resolving URSCs at mitotic onset.

To understand better the mechanism of URSC and mitotic DNA bridge formation, we used drug treatments to induce replication stress via various mechanisms in control and traip⁻ larvae and assayed their survival to adult (Fig. 6C). A major source of URSCs is late-replicating regions, where replication forks fail to converge before mitotic entry (Liu et al., 2014); hydroxyurea increases late-replicating regions by depleting the nucleotide pool and inhibiting replication (Bianchi et al., 1986). Surprisingly, traip⁻ mutants were not sensitive to hydroxyurea treatment (Fig. S4A), suggesting that late-replicating region-induced URSCs are not the major target for Traip.

Another source of URSCs is inter-strand crosslinks, which block replication machinery and cause stalled forks; both cisplatin and mitomycin C are inter-strand crosslinking agents (Deans and West, 2011). Whereas controls tolerated high doses of cisplatin and mitomycin C, traip⁻ animals only tolerated low doses, with higher doses causing lethality (Fig. 6D), likely due to defects in imaginal disc proliferation (Fig. S4B). The MBs of traip⁻ adults raised on cisplatin or mitomycin C were not affected (Fig. S4C,D); however, this may be due to a failure of these drugs to cross the blood brain barrier, as is the case in mammals (Gregg et al., 1992; Reddy and Randerath, 1987). Nonetheless, these experiments indicate that Traip functions to resolve DNA crosslink-induced damage.

TRAIP triggers fork unloading via ubiquitylation of MCM7 (Wu et al., 2019), and therefore requires an intact RING domain for E2-conjugating enzyme binding. Thus, we tested whether the RING domain is required for Traip function in MB development. GFP::traip^L12E,G14D, which is a control mutant predicted to have a functional RING domain, fully rescued traip⁻ MB lobe size (Fig. 6E,F). In contrast, GFP::traip^I8D,A10E, which is predicted to have a non-functional RING-E2 binding interface, failed to rescue traip⁻ MB lobe size (Fig. 6F,G). GFP::traip^ΔRING enhanced the traip⁻ MB lobe size defect (Fig. 6H), indicating a dominant-negative effect of deleting the RING domain. These results show that traip function in MB-NBs is RING dependent, consistent with the known E3 ligase function and role in fork unloading for TRAIP.

We identified for the first time mitotic localization for Traip while characterizing mNG::Traip expression. In interphase, mNG::Traip localized to the nucleus, as previously described (Feng et al., 2016; Harley et al., 2016; Soo Lee et al., 2016). However, during mitosis mNG::Traip became localized to mitotic spindles and concentrated at the centrosomes (Fig. 7A, Movie 5). High-resolution imaging revealed that mNG::Traip forms small puncta that travel poleward along the spindles (Fig. 7B, Movie 6) and also coalesce at the cytokinetic furrow and midbody (Fig. 7C) in late mitosis.

To characterize regions within Traip responsible for its mitotic localization, we expressed GFP::Traip variants in traip⁻ CB-NBs via wor-GAL4, including full length (FL), RING mutant (I8D, A10E), and several truncations and internal deletions based on the major features of Traip (Fig. S5A). GFP::Traip^FL recapitulated the interphase and mitotic localizations of mNG::Traip (Fig. 7D, Fig. S5B). GFP::Traip^I8D,A10E localized to the proper sites, but formed aggregates (Fig. S5C). Neither the RING domain nor the first coiled-coil domain was sufficient to mediate any localization (Fig. S5D). Both the second coiled-coil domain and the C-terminal region were sufficient to mediate centrosome localization (Fig. S5E-H). Finally, the C-terminal region was necessary and sufficient for localization to the furrow during mitosis and the nucleus/nucleolus during interphase (Fig. S5G,H). We also tested whether these protein regions are required for Traip MB function, but none was sufficient to rescue traip⁻ MB lobe size (Fig. S6A), indicating that Traip function requires multiple domains. Human GFP::TRAIP also failed to rescue traip⁻ MB lobe size (Fig. S6A); although the domain structures of Traip and TRAIP are conserved, they are fairly divergent at the protein level (22% identity; Fig. S6B).

We next tested whether mitotic localization of Traip was sufficient for proper MB development. GFP::Traip^ΔNLS, which contains a deletion of the nuclear localization signal (NLS), localized similarly to wild-type GFP::Traip during mitosis but was cytoplasmic in interphase (Fig. 7F). Nonetheless, GFP::traip^ΔNLS fully rescued traip⁻ adult MB lobe size (Fig. 7G,H), 24 h APF MB-NB number (Fig. 7I), KC number (Fig. 7J,K) and nearly completely rescued mitotic chromosome bridging (22/23, Fig. 7L, Movie 7), indicating that Traip function during mitosis is sufficient for MB-NBs. To test whether proper mitotic localization is important for Traip function, we introduced a Tom20 tag to ectopically force Traip to the mitochondria. Tom20::TagRFP::Traip localized cytoplasmically in interphase, and was absent from the spindle region in mitosis (Fig. 7M). Mitochondria-localized Tom20::TagRFP::Traip provided an intermediate rescue of traip⁻ MB lobe size (Fig. 7H,N). Given the centrosome and spindle localization and the apparent importance of proper mitotic localization, we hypothesized that traip⁻ MB-NBs could have centrosome or spindle defects that could contribute to microcephaly phenotypes. However, we detected no abnormalities in centrosome numbers, γ-Tubulin recruitment to centrosomes (Fig. S7A), or in the polarized localization of aPKC (Fig. S7B) in metaphase 24 h APF MB-NBs. Thus, proper mitotic localization of Traip to the spindle, centrosome and/or midbody is important for full Traip function.

Our study in Drosophila shows that traip⁻ shares several characteristics with human microcephaly mutants. First, the traip⁻ phenotype is highly brain specific, with body defects being rare. Second, the traip⁻ MB phenotype is developmental rather than neurodegenerative, reflecting a primary rather than secondary microcephaly-like disorder. Finally, as with many human microcephaly genes, traip functions to promote NPC proliferation and survival. Thus, traip⁻ represents a powerful new disease model for understanding the etiological mechanisms underlying microcephaly.

Despite their ubiquitous expression, mutations in microcephaly genes primarily affect the cerebral cortex in humans. Similarly, both traip and the DDR microcephaly gene MCPH1 (Rickmyre et al., 2007) are ubiquitously expressed in Drosophila, yet the MB is the only adult structure affected in their mutants. Although many tissues can make up for lost cells via compensatory proliferation (Haynie and Bryant, 1977; Pfau et al., 2016), no such process appears to exist for replacing lost NPCs. Additionally, whereas most NBs have a limited window of proliferation, MB-NBs divide continuously from embryogenesis into late pupal stages (Ito and Hotta, 1992; Truman and Bate, 1988), potentially allowing more accumulation of rare or small effects over many cell cycles. Thus, we speculate that mutations in microcephaly genes likely affect all CB-NBs to some degree, but the MB-NBs are especially sensitive to these mutations as a consequence of their relatively prolonged period of proliferation. We speculate that a similar explanation, including a prolonged period of rapid proliferation and lack of compensatory proliferation, may account for the sensitivity of the human cortex to microcephaly gene mutation.

Our work provides the first link between a known function of Traip and proper brain development. We found that interphase nuclear localization is not required for Traip function, suggesting that Traip interphase functions are dispensable for MB-NB survival under normal conditions. Instead, we discovered the presence of mitotic DNA bridges, sensitivity to inter-strand crosslinking agents, and RING domain dependence, consistent with the well-established role of TRAIP in unloading stalled forks to initiate repair (Fig. 8A; Deng et al., 2019; Priego Moreno et al., 2019; Sonneville et al., 2019). Furthermore, GFP::Traip^ΔNLS rescue experiments suggest either that Traip primarily performs this unloading function during mitosis (Deng et al., 2019), or else that Traip normally functions during interphase but is able to unload stalled forks during mitosis if necessary. Alternatively, we cannot rule out the possibility that there may be residual GFP::Traip^ΔNLS in the nucleus that allows continued function during interphase, or else that nuclear localization of Traip is not required for an interphase function. We surmise that traip⁻ MB-NBs have stalled replication machinery that remains loaded throughout mitosis, preventing mitotic DNA synthesis repair and proper sister chromatid segregation (Fig. 8B). As anaphase proceeds, attached sister chromatids are pulled to opposite poles and they form UFBs as the under-replicated DNA is stretched out between them. These bridges could be physically broken, leading to chromosome fragmentation, generating aneuploidy or micronuclei and causing nuclear deformations in daughter cells (Gisselsson et al., 2001; Heddle and Carrano, 1977). Alternatively, persistence of DNA bridges at the cytokinetic furrow could induce mitotic exit and furrow regression (Pampalona et al., 2012; Shi and King, 2005), leading to multiple nuclei or polyploidy, which likely prevent further proliferation.

Under normal conditions, MB-NBs are lost at the end of pupal development via caspase-dependent apoptosis (Siegrist et al., 2010). Here, we found that traip⁻ MB-NBs are lost prematurely, in part via caspase-dependent cell death, and thus fail to generate proper KC numbers. However, our caspase-inhibition experiments did not fully suppress traip⁻ MB phenotypes, suggesting that additional redundant mechanisms may play a role in MB-NB loss. For example, when caspase-dependent apoptosis is inhibited, MB-NBs are primarily lost via autophagy (Pahl et al., 2019). Alternatively, the irregular, crenellated nuclear envelope morphology of some traip⁻ MB-NBs (Fig. 5C) could point to non-apoptotic cell death pathways (Kutscher and Shaham, 2017). Finally, aneuploidy-induced cell cycle exit in traip⁻ MB-NBs could lead to loss via premature differentiation (Gogendeau et al., 2015). Furthermore, it is likely that loss of KCs and/or GMCs also contributes some to traip⁻ MB size defects.

DNA bridge-induced defects likely feed into premature cell loss, but further work is required to dissect the pathways connecting them. In Drosophila, polyploid NBs can accumulate significant DNA damage as they enter mitosis (Nano et al., 2019), and chromosome breakage during mitosis in traip⁻ could induce death through DNA-damage signaling. Drosophila embryos laid by traip⁻ mothers do not survive, with extensive chromosome bridging and Chk2-dependent cell death, suggesting that DNA damage accumulation leads to cell loss in the rapidly dividing cells of the early embryo (Merkle et al., 2009). In mammalian NPCs, polyploidy and binucleation can cause G1 arrest and apoptosis (Aylon and Oren, 2011; Storchova and Kuffer, 2008). In Drosophila, neurons can become polyploid in response to DNA damage (Nandakumar et al., 2020), and NBs can become massively polyploid in some mutants (Poulton et al., 2017; Swider et al., 2019), suggesting that, even though polyploidy may be better tolerated in flies, polyploid NBs are unlikely to complete additional mitoses successfully. We infer the existence of traip⁻ aneuploid MB-NBs, which produce a wide range of daughter KC numbers, suggesting that traip⁻ generates some aneuploidies that are well tolerated and others that are highly lethal. Similarly, one recent study found that, although many karyotypes are permitted in NBs, loss of both copies of any of the three major Drosophila chromosomes resulted in a failure to proliferate and likely elimination (Mirkovic et al., 2019). This parallels the situation in mammals, in which aneuploid NPCs and neurons are common (Rehen et al., 2001), but also sensitive to G1 arrest, cell cycle exit, and apoptosis (Peterson et al., 2012; Storchova and Kuffer, 2008). Thus, both polyploidy and aneuploidy could stop further proliferation in traip⁻ MB-NBs by preventing proper mitosis or inducing G1 arrest and cell cycle exit, eventually triggering cell loss via various mechanisms.

In this study, we identified centrosome, spindle and cytokinetic furrow localizations for Traip that are important for function. One possibility is that the dynamic movement of Traip on the mitotic spindle and cytokinetic furrow promotes encounters with unresolved DNA bridges. We never observed GFP::Traip on bridges. However, as a single TRAIP protein is probably sufficient to unload each replisome (Wu et al., 2019), fluorescence detection may be unlikely. Interestingly, centrosome localization is a common aspect of microcephaly-linked proteins, including MCPH1, which also functions in DDR (Jeffers et al., 2008; Rai et al., 2008). Similar to Traip, MCPH1 has mitotic functions required for proper chromosome segregation, and mutations in MCPH1 lead to lagging chromosomes, DNA bridges and micronuclei (Arroyo et al., 2017; Rickmyre et al., 2007). Mutations in microcephaly genes with centrosome-associated functions, such as CEP135 and CDK5RAP2, cause dysregulation of centrosome numbers, which also lead to chromosome segregation errors and aneuploidy (Barrera et al., 2010; Hussain et al., 2012; Shi et al., 2019). Thus, mitotic roles, ensuring proper chromosome segregation, and suppressing aneuploidy are common features of microcephaly-linked proteins. Future work seeking to understand these shared defects better may reveal a deeper etiological connection across microcephaly disorders.

Details of reagents, strains and resources used are in Table S1.

Experimental fly crosses were maintained on Bloomington Recipe Fly food from LabExpress (Ann Arbor, MI, USA) and kept at 25°C. Crosses were either eight virgin females per vial or 20 virgin females per bottle, with at least half as many males. yw was used as a control in all genotypes marked ‘+’. All new transgenic animals were generated using standard embryo injection protocols by BestGene (Chino Hills, CA, USA).

For traip^Δ, homology arms were cloned into pUC-attP-3xP3-TR, such that the repair construct contained 1100 bp upstream of the traip start codon and 1094 bp downstream of the traip stop codon flanking a attP-3xP3-TR replacement. For mNG::traip, homology arms and mNeonGreen were cloned into pUC57, such that the repair construct contained 2226 bp upstream of the traip start codon, the mNeonGreen coding sequence, and 1093 bp downstream of the traip start codon. Guide RNAs were cloned into pU6-chiRNA. Note that PAM sequences in homology arms were mutated such that they were not predicted to be targets for Cas9. Repair template and gRNA plasmid constructs were injected into y¹ M{RFP[3xP3.PB] GFP[E.3xP3]=vas-Cas9}ZH-2A w¹¹¹⁸/FM7c using standard procedures. Putative CRISPR-positive alleles were crossed to CyO-Cre to eliminate 3xP3-TagRFP and were fully sequenced from outside the homology arms. traip^Δ was further back-crossed to yw for three generations and re-sequenced to eliminate a second site lethal mutation. Both CRISPR alleles were crossed to Df(2R)Exel7153 to test for maternal effect lethality; traip^Δ was maternal-effect lethal, and mNG::traip was fertile.

traip^I8D,A10E and traip^L12E,G14D were rationally designed using amino acid alignment of Traip and TRAF6, which is a RING domain E3 ligase that has a known crystal structure with its E2-conjugating enzyme Ubc13 (Yin et al., 2009). traip^ΔNLS was designed using cNLS Mapper, which predicts importin α-dependent, CDK1-regulated NLS sequences (Kosugi et al., 2009). traip^I8D,A10E, traip^L12E,G14D and traip^ΔNLS (deletion of positions 357-366 YSIFKKPRLL) were generated using mutagenic primers. traip truncations were PCR-amplified from cDNA. traip, traip variants, human TRAIP and FancD2 cDNAs were cloned into pENTR using standard Gateway cloning methods. Gateway cloning was used to move constructs from pENTR into pPGW for UAS-GFP::traip^variants, pUGW for ubi-GFP::traip, pPT20TRW for UAS-Tom20::TagRFP::traip, and pUNW for ubi-mNG::FancD2.

Parent flies (traip^Exc142/CyO females and either yw or traip^Δ males) were crossed for 2-3 days and then transferred to fly food made from 0.6 g of Carolina Formula 4-24 (Fisher Scientific) and 2 ml of ddH₂O. Parents laid eggs for 24 h, eggs were allowed to hatch for 24 h, and then 450 μl of yeast solution (1 g dry yeast sprinkles per 10 ml ddH₂O) with drug was added on top of the food. Drugged larvae were allowed to develop and adults were scored by genotype daily as they eclosed. Concentrated drug stock solutions were reconstituted as follows: Cisplatin at 1 mg/ml in PBS; mitomycin C at 0.5 mg/ml in ddH₂O; hydroxyurea at 5 mg/ml in ddH₂O.

Control, traip⁻ and traip⁻+ubi-GFP::traip flies were processed for μ-CT according to Schoborg et al. (2019). Briefly, adult flies were aged to 2-4 days old, anesthetized on CO₂, de-waxed in PBST (0.5% Triton X-100 in PBS), and fixed in Bouin's solution for 24 h. Fixed flies were washed in 0.1 M Na₂HPO₄ with 1.8% sucrose and then stained in 0.1 N solution of I₂KI for 2 days. Stained flies were washed with ultrapure H₂O and scanned using SkyScan 1172 desktop scanner. Tomograms were generated with NRecon, and 3D volumetric segmentation was performed using Dragonfly. Brain measurements were normalized to thorax width to control for overall size differences among individual flies.

Animals at appropriate stages were collected and aged: adult flies were aged to 3-4 days old unless otherwise stated; white pre-pupae were collected and aged for 24, 48 or 72 h to obtain pupal stages. Staining procedures were modified from Jenett et al. (2012) and Aso et al. (2014). Tissues were dissected in SF900 S2 cell media (Fisher Scientific) and fixed with 2% paraformaldehyde in SF900 for 2 h at room temperature (RT) or overnight at 4°C. Fixed samples were washed three times in PBST and blocked for 1.5 h in PBST+5% normal goat serum (NGS). Blocked samples were incubated in PBST+NGS with primary antibody for 1-2 h at RT followed by one or two overnight incubations at 4°C. Primary antibody concentrations were as follows: mouse anti-FasII [Developmental Studies Hybridoma Bank (DSHB), 1D4; 1:50]; rat anti-CadN (DSHB, DN-Ex#8; 1:25); mouse anti-Nrg (DSHB, BP104; 1:25); mouse anti-Elav (DSHB, 7E8A 10; 1:100); mouse anti-Lamin (DSHB, ADL84.12; 1:100); mouse anti-Tubulin (DSHB, E7; 1:200); rat anti-DCAD2 (DSHB, DCAD2; 1:20); rabbit anti-γH2Av (Rockland, 600-401-914; 1:100); mouse anti-GFP (Thermo Fisher, A11122; 1:1000); mouse anti-phospho-tyrosine (Millipore, 4G10; 1:100); rabbit anti-phospho-histone H3 (Millipore, H3S10P; 1:40,000); rabbit anti-Dcp-1 (Cell Signaling Technology, 9578; 1:500); mouse anti-γTubulin (Sigma-Aldrich, GTU-88; 1:500); guinea pig anti-Asl (Klebba et al., 2013; 1:20,000); rabbit anti-aPKC (Santa Cruz Biotechnology, C20; 1:1000); guinea pig anti-Dpn (kind gift of J. B. Skeath, Washington University in St Louis, MO, USA; 1:50). Samples were washed five times in PBST+NGS, and then incubated in PBST+NGS with 1:500 secondary antibody (Thermo Fisher, various goat anti-mouse/rabbit/rat/guinea pig secondary antibodies conjugated to Alexa fluor 488/568/647: A11001, A11004, A21235, A21124, A21241, A21242, A11008, A11011, A21244, A11077, A21247 and A21450) and 1× DAPI (Invitrogen, D21490) for 1-2 h at RT followed by one to three overnight incubations at 4°C. Samples were washed three times in PBST+NGS and then three times in PBST, before being post-fixed in 4% paraformaldehyde in PBS for 1 h at RT. Post-fixed samples were washed twice in PBST, and then washed in PBS until all Triton X-100 was removed. Samples were mounted on poly-L-lysine-coated #1.5 coverslips, rinsed in ddH₂O, and then dehydrated through a series of 30%, 50%, 75%, 95% and then three 100% ethanol 10 min baths. Dehydrated coverslips were then bathed three times in 100% xylene for 5 min, before being mounted in DPX on slides with #1.5 coverslips as spacers. DPX was cured at RT overnight. For phalloidin-stained samples, Phalloidin–Atto 647N (Sigma-Aldrich) was added at 1:40 in PBS for 20 min after secondary antibody incubation, and samples were not post-fixed and were mounted in AquaPoly Mount (Polysciences).

Most fixed samples were imaged using a Zeiss LSM 880 confocal microscope with a 63×/1.4 NA objective, GaAsP detectors, and 405, 488, 561 and 641 nm laser lines, controlled using Zen Black software. Some fixed samples (Figs 4A, 5A,I,J) were imaged using a Nikon W1 spinning disk confocal equipped with a Prime BSI cMOS camera (Photometrics) and a 100×/1.4 NA silicon immersion objective, controlled using Nikon Elements software. Live imaging was performed either using the same Nikon W1 with 100×/1.4 NA silicon or 40×/1.3 NA oil immersion objective, or using an Eclipse Ti2 (Nikon) with a 100×/1.49 NA objective or a 40×/1.3 NA objective and a 1.5× tube lens, an ORCA-Flash 4.0 CMOS camera (Hamamatsu Photonics), and 405, 491, 561 and 642 nm laser lines, controlled by MetaMorph software. For live imaging, larval brains were dissected in Schneider's media and mounted by sandwiching between a #1.5 coverslip and a 50-mm lummox dish, with droplets of Halocarbon oil 700 as a cushion and surrounding the edge of the coverslip.

For MB α lobe cross-sectional analysis, image stacks of α lobes were stack rotated to be perpendicular to the viewing plane using the Interactive Stack Rotation plugin in ImageJ. Rotated MB lobes were then cross-sectioned through the middle of the lobe using the reslice tool, and the area of the resliced MB lobe was measured in ImageJ. For MB volumetric analysis, either FasII-positive or OK107-GAL4>mCD8::GFP-positive MB lobes were segmented and measured using the Pixel Classification and 3D Object Analysis tools in Aivia software. For KC counting analysis, the OK107-GAL4>NLS::mCherry-positive KCs were similarly segmented and counted using Pixel Classification and 3D Object Analysis tools in Aivia software. For Dcp-1-positive KC normalization, total KCs were estimated via manual counting. γH2Av puncta were thresholded and identified using the 3D Objects Counter in ImageJ. Then, γH2Av objects inside the nucleus, as marked by anti-Lamin immunostaining, were counted in each MB-NB. ImageJ was used to generate maximum intensity projections for presentation.

Data analysis was performed using Microsoft Excel and GraphPad Prism. In all graphs, the mean±s.d. and all individual data points are presented. Sample sizes were primarily determined based on the availability of animals of the proper genotype and developmental stage, and the time required for dissection, processing and imaging; given these considerations, we processed 8-12 brains per genotype for most experiments, although some brains were not imaged owing to physical damage. Experiments measuring MB lobe size were typically performed once, with control and traip^Δ conditions repeated in each experiment. Shapiro–Wilk test was used to test the assumption of normality. Statistical tests used to determine significance are reported in figure legends.

We thank the National Heart, Lung, and Blood Institute Light Microscopy Core, especially Xufeng Wu, for support with confocal microscopy. We thank Bloomington Drosophila Stock Center for fly stocks, and Developmental Studies Hybridoma Bank for antibodies. Carey Fagerstrom performed all cloning. Rachel Ng performed the μ-CT experiments. We thank Alex Kelly, Matthew Hannaford and Ramya Varadarajan for helpful discussion, and Todd Schoborg, Brian Galletta, Ed Giniger and Hong Xu for critically reading the manuscript.

The peer review history is available online at https://journals.biologists.com/dev/article-lookup/doi/10.1242/dev.199987.