Parafibromin governs cell polarity and centrosome assembly in Drosophila neural stem cells

Neural stem cells (NSCs) divide asymmetrically to balance their self-renewal and differentiation, an imbalance in which can lead to NSC overgrowth and tumor formation. The functions of Parafibromin, a conserved tumor suppressor, in the nervous system are not established. Here, we demonstrate that Drosophila Parafibromin/Hyrax (Hyx) inhibits ectopic NSC formation by governing cell polarity. Hyx is essential for the asymmetric distribution and/or maintenance of polarity proteins. hyx depletion results in the symmetric division of NSCs, leading to the formation of supernumerary NSCs in the larval brain. Importantly, we show that human Parafibromin rescues the ectopic NSC phenotype in Drosophila hyx mutant brains. We have also discovered that Hyx is required for the proper formation of interphase microtubule-organizing center and mitotic spindles in NSCs. Moreover, Hyx is required for the proper localization of 2 key centrosomal proteins, Polo and AurA, and the microtubule-binding proteins Msps and D-TACC in dividing NSCs. Furthermore, Hyx directly regulates the polo and aurA expression in vitro. Finally, overexpression of polo and aurA could significantly suppress ectopic NSC formation and NSC polarity defects caused by hyx depletion. Our data support a model in which Hyx promotes the expression of polo and aurA in NSCs and, in turn, regulates cell polarity and centrosome/microtubule assembly. This new paradigm may be relevant to future studies on Parafibromin/HRPT2-associated cancers.

Human Parafibromin/Cell division cycle 73 (Cdc73)/hyperparathyroidism type 2 (HRPT2) is a tumor suppressor that is linked to several cancers, including parathyroid carcinomas and hyperparathyroidism-jaw tumor syndrome, head and neck squamous cell carcinomas, as well as breast, gastric, colorectal, and lung cancers [45][46][47][48]. Somatic mutations in parafibromin have been found in 67% to 100% of sporadic parathyroid carcinomas [45]. Parafibromin is part of a conserved polymerase-associated factor complex that primarily regulates transcriptional events and histone modification [49,50]. Hyrax (Hyx), Drosophila Parafibromin, is essential for embryonic and wing development and is known to positively regulate Wnt/Wingless signaling pathway in wing imaginal discs by directly interacting with β-catenin/Armadillo [51]. Human Parafibromin, but not yeast Cdc73, rescues defects in wing development and the embryonic lethality caused by hyx loss-of-function alleles [51], suggesting that Parafibromin functions during development are conserved across metazoans. Interestingly, Parafibromin is expressed in both mouse and human brains, including the cortex, basal ganglia, cerebellum, and the brainstem [52], suggesting that Parafibromin may play a role in central nervous system (CNS) functions. However, the specific functions of Parafibromin in the nervous system are not established. Here, we investigate the role of Parafibromin/Hyx in the asymmetric division of NSCs during Drosophila larval brain development.

Loss of hyx results in NSC overgrowth in the larval central brain
In a clonal screen of a collection of chromosome 3R mutants induced by ethyl methanesulfonate (EMS) [53], we identified 2 new hyrax (hyx) alleles-hyx HT622 and hyx w12-46 -which produce NSC overgrowth phenotype in the central brain of Drosophila larvae at 96 h after larval hatching (ALH) (Fig 1A and 1B). hyx/CG11990 encodes a highly conserved, 531-amino acid protein that is homologous to mammalian Parafibromin/Cdc73. The hyx HT622 allele contains a 74-bp deletion from nucleotides 728 (immediately after amino acid 242) to 801, which results  G) and 1456 (T to A), which causes amino acid substitutions-Leucine (L) to Arginine (R) at amino acid 444 and Cysteine (C) to Serine (S) at amino acid 486, respectively. As hyx HT622 and hyx w12- 46 homozygotes are embryonically lethal, we generated MARCM (mosaic analysis with a repressible cell marker) clones [54] to examine the clonal phenotype at larval stages. Hyx protein in the clones was detected using guinea pig anti-Hyx antibodies that we generated against the N-terminal 1-176 amino acids of Hyx. In wildtype control clones, Hyx was predominantly localized to the nuclei of NSCs (S1A and S1B Fig;  0.74-fold in the nuclei, 0.26-fold in the cytoplasm, total intensity:1; n = 15 NSC) and their progeny, but not on mitotic structures (S1A and S1B Fig). In contrast, Hyx was undetectable in 91.3% and 40% of clones generated from hyx HT622 and hyx w12-46 alleles, respectively, and was dramatically reduced in the rest of the clones. The intensity of Hyx was significantly reduced to 0.15-fold in hyx HT622 NSCs (n = 25) and 0.32-fold in hyx w12-46 NSCs (n = 28), in contrast to 1-fold in control NSCs (S1A-S1C Fig; n = 20). Moreover, we performed western blot with protein extracts from FRT82B, hyx HT622 , and hyx w12-46 homozygous embryos, as hyx homozygous mutants do not survive to larval stages. Relative protein intensity of Hyx at 24 h after egg laying (AEL) was reduced to 0.31-fold and 0.26-fold in both hyx HT622 and hyx w12- 46 , respectively (S1D and S1E Fig; control, 1-fold). Maternal Hyx might partially contribute to the detected Hyx proteins in these samples. Hyx levels were also significantly reduced in hyx RNAi (RNA interference) in hyx HT622 /+ background to 0.22-fold in third instar larval brains driven by NSC-specific driver insc-Gal4 (S1D and S1E Fig). This western blot result is consistent with our immunofluorescence data that Hyx is dramatically diminished with weak signal in these hyx mutants at the third instar stage (S1A-S1C Fig). Therefore, these observations indicate that hyx HT622 and hyx w12-46 are 2 strong loss-of-function alleles.
In the central brain of Drosophila larvae, there are at least 2 types of NSCs, both of which divide asymmetrically [55][56][57]. Each type I NSC generates another NSC and a ganglion mother cell (GMC) that gives rise to 2 neurons, while each type II NSC produces an NSC and a transient amplifying cell (also known as an intermediate neural progenitor or INP), which, in turn, go through a few cycles of asymmetric divisions to produce GMCs [55][56][57][58]. In the wild-type control, only 1 NSC is maintained in each of type I or type II MARCM clones (Fig 1B-1D). However, ectopic NSCs were observed in 87% of type I clones and 75% of type II clones generated from the hyx HT622 allele (Fig 1B). Similarly, supernumerary NSCs were observed in both type I and type II NSC lineages from hyx w12-46 clones (Fig 1B). Ectopic NSCs per clone were observed in type I hyx HT622 (5.7) and hyx w12-46 (2.8) clones and type II hyx HT622 (6.2) and hyx w12-46 (5.0) clones ( Fig 1D). In addition, knockdown of hyx by 2 independent RNAi lines, under the control of an NSC driver insc-Gal4, led to the formation of multiple NSCs in both type I and type II lineages (S1F and S1G Fig). Moreover, the NSC overgrowth phenotype in hyx HT622 and hyx w12-46 mutants was fully rescued by the overexpression of a wild-type hyx transgene (S2A and S2B Fig). Therefore, our finding shows that hyx prevents NSC overgrowth in both type I and type II lineages.
Next, we wondered whether the supernumerary NSCs detected in hyx-depleted larval brains could persist in proliferation in pupal stages. At 10 h after puparium formation (APF), 88.2% of the NSCs in pupal brains were negative for PH3 (Fig 1E and 1F). In contrast, 54.8% of hyx RNAi hyx HT622/+ NSCs were proliferative and marked by PH3, suggesting that increased number of hyx-depleted NSCs were still actively dividing in early pupal stages. This phenotype might be due to an NSC decommissioning defect or abnormal NSCs that were generated by symmetric divisions acquired proliferative potential. Temporal transcription factors are known to schedule Pros-dependent cell cycle exit of NSCs at the end of larval stages [59]. The abnormal hyx-deficient NSCs often undergo symmetric division, which might result in a disruption of temporal factor transition and, in turn, continued proliferation after pupal formation.
Remarkably, the ectopic NSC phenotype observed in hyx HT622 larval brains was completely rescued by the overexpression of Parafibromin/HRPT2, the human counterpart of Hyx (S2C Fig). Likewise, the NSC overgrowth phenotype observed in hyx knockdown brains was fully restored by the introduction of human Parafibromin/HRPT2 in both type I and type II NSC lineages (S2D Fig). Therefore, Parafibromin/Hyx appears to have a conserved function in suppressing NSC overgrowth.
Parafibromin/Cdc73 (Hyrax/Hyx in Drosophila) is a component of the Paf1 complex, an evolutionarily conserved protein complex that functions in gene regulation and epigenetics [60][61][62][63]. The Paf1 complex also consists of other core subunits Paf1 (Antimeros/atms in Drosophila), Leo1 (Another transcription unit/Atu in Drosophila), Cln three requiring 9 (Ctr9 in Drosophila), and Rtf1 [64,65]. We sought to analyze the function of other components of the Paf1 complex in the larval central brains. Surprisingly, although Ctr9 is required to terminate the proliferation of Drosophila embryonic NSCs [33,66], no ectopic NSCs were observed in ctr9 12P023 type I and type II MARCM clones (S2E Fig). Similarly, knockdown of ctr9, atms, or atu under the control of insc-Gal4 did not generate supernumerary NSCs in either type I or type II lineages in the larval central brains (n = 5 for all). Interestingly, knocking down rtf resulted in a weak ectopic NSC phenotype in type II lineages, without affecting type I NSC lineage development (rtf1 RNAi/BDSC#34586: type II, 31.2%, n = 64 and rtf1 RNAi/BDSC#34850: type II, 7%, n = 43). Therefore, our results suggest that Parafibromin/Hyx might prevent NSC overgrowth independent of Paf1 complex function during Drosophila brain development.

Hyx is essential for the asymmetric division of NSCs
The generation of ectopic NSCs in the absence of Hyx function was not due to INP dedifferentiation, as no type II NSCs were generated in both control and hyx RNAi/V103555 derived INP clones (S3A Fig). hyx expression was efficiently down-regulated in these INP clones upon hyx knockdown (S3B Fig). Next, we assessed whether hyx is required for the asymmetric division of NSCs. In wild-type control metaphase NSCs, apical proteins such as aPKC, Insc, Baz/ Par3, Par6, and Pins were localized asymmetrically in the apical cortex (Fig 2A-2F). By contrast, aPKC in hyx HT622 and hyx w12-46 metaphase NSCs was completely delocalized from the apical cortex to the cytoplasm (Fig 2A and 2B). Similarly, other apical proteins including Insc, Baz, Par6, and Pins in hyx HT622 metaphase NSCs were no longer localized asymmetrically in the apical cortex, exhibiting weak or punctate signals in the cytoplasm (Fig 2C-2F).
Next, we examined the localization of basal proteins in hyx mutant NSCs. Mira was asymmetrically localized in the basal cortex in 100% of wild-type NSCs during metaphase, but its basal localization was severely disrupted in metaphase NSCs of hyx HT622 and hyx w12-46 clones (Fig 2A and 2B). Similarly, Numb and Brat lost their asymmetric basal localization and were observed in the cytoplasm in metaphase NSCs of hyx HT622 clones (Fig 2C and 2G).
Similarly, hyx knockdown by 2 independent RNAi lines disrupted NSC apicobasal polarity. aPKC, Par6, Baz, Insc, and Pins were delocalized from the apical cortex in all metaphase NSCs To examine the daughter cell size asymmetry, we measured the ratio of NSC to GMC diameter at telophase. Cells were outlined by a membrane marker CD8-GFP driven by insc-Gal4. The NSC-to-GMC diameter ratio in hyx RNAi telophase NSCs was significantly reduced to 1.2 (Fig 2H and 2I; n = 4) compared with 1.8 (n = 11) in control telophase NSCs. These data indicate that hyx depletion disrupts daughter cell size asymmetry of the NSCs. Given that polarization of NSCs is an essential prerequisite for their asymmetric division, we sought to investigate whether hyx depletion could result in the symmetric division of NSCs. To this end, we took advantage of a microtubule-binding protein Jupiter-GFP (also known as G147-GFP), which is controlled under its endogenous promoter [67]. In live, whole-mount larval brains that expressed G147-GFP, control NSCs always divided asymmetrically to produce 2 daughter cells with distinct cell sizes ( Fig 2J and S1 Movie). By contrast, all NSCs in hyx RNAi divided symmetrically to generate 2 daughter cells with similar cell sizes ( Fig 2J and S2 Movie). These observations indicate that hyx-depleted NSCs divide symmetrically, leading to NSC overgrowth. However, the allograft transplantation of hyx HT622 (n = 29 host) and hyx w12-46 (n = 30 host) homozygous clones and hyx RNAi (n = 30 host) larval brain cells did not induce tumors. Perhaps hyx-depleted cells are small in size and with altered cell fate and have reduced cell growth, thus unable to expand in this tumor assay.
The protein levels of the phosphorylated Akt (Ser505) (P-Akt) were unaffected in hyx RNAi NSCs as compared with that in control NSCs (S4A and S4B Fig; 1

PLOS BIOLOGY
hyx HT622 and hyx w12-46 NSCs were much shorter than that in the control (Fig 3C and 3D). When normalized against the cell diameter of NSCs, the mitotic spindle length was still significantly shortened in hyx-depleted NSCs. The relative spindle length was significantly shortened to 0.74-fold in hyx RNAi NSCs (Fig 3E; n = 15 NSC) compared with 0.82-fold in control NSCs (n = 32 NSC). Similarly, the ratio of spindle length to cell diameter was significantly reduced to 0.70-fold in both hyx HT622 (n = 13 NSC) NSCs and hyx w12-46 NSCs (n = 21 NSC) when compared with 0.79-fold in control NSCs (Fig 3F; n = 22 NSC).
These observations prompted us to examine whether Hyx is important for microtubule assembly in NSCs. We sought to determine whether Hyx regulates the formation of microtubule asters in interphase NSCs. A wild-type interphase NSC forms 1 major microtubule aster marked by α-tubulin (α-tub). Asters are assembled by the microtubule-organizing center (MTOC), also known as centrosomes, of cycling NSCs labeled by a centriolar protein called Asterless (Asl; Fig 3G). Strikingly, the vast majority of interphase NSCs in hyx HT622 and hyx w12-46 clones either failed to organize a microtubule aster or formed weak microtubule asters ( Fig 3G). The astral microtubule intensity marked by α-tub was dramatically reduced to 14.7 (a.u.) and 27.6 (a.u.) in hyx HT622 (n = 23 NSC) and hyx w12-46 NSCs (n = 23 NSC), respectively, significantly lower than 100.7 (a.u.) in control NSCs (Fig 3H; n = 7 NSC). Likewise, in hyx knockdown clones, 50% of NSCs failed to form microtubule asters and 40.9% of NSCs only assembled weak microtubule asters during interphase (S5A Fig). Overall, we show that Hyx is important for the formation of interphase microtubule asters.
The shortened mitotic spindles and defects in the assembly of microtubule asters upon hyx depletion suggested that Hyx might regulate microtubule growth in dividing NSCs. To this end, we performed a microtubule regrowth assay by "cold" treatment of larval brains on icefor efficiently depolymerizing microtubules in NSCs-followed by their recovery at 25˚C to allow microtubule regrowth in the course of time. In both control and hyx RNAi interphase NSCs treated with ice (t = 0), no astral microtubules were observed and only weak residual microtubules labelled by α-tub remained at the centrosome ( Fig 4A). The centrosomes in 76.3% of these hyx RNAi interphase NSCs were absent or much smaller in size, suggesting that the MTOC was compromised upon hyx depletion ( Fig 4A). In control interphase NSCs, robust astral microtubules were observed around the centrosome, at various time points following recovery at 25˚C (Fig 4A). By contrast, the vast majority of hyx RNAi interphase NSCs reassembled scarce microtubule bundles without detectable MTOCs, even 120 s after recovery at 25˚C ( Fig 4A).
Next, we examined microtubule regrowth in mitotic NSCs. Upon treatment with ice (t = 0), spindle microtubules were effectively depolymerized with residual microtubules marking the centrosomes/spindle poles, in all control and hyx RNAi metaphase NSCs (Fig 4B). Consistent with poor centrosome assembly during interphase, the centrosomes of 98.0% of hyx RNAi metaphase NSCs were deformed with irregular shapes (Fig 4B). In all control metaphase NSCs, intense spindle microtubules were reassembled around centrosomes and chromosome mass from as early as 30 s following recovery at 25˚C; the mitotic spindle completely reformed at 2 min following recovery ( Fig 4B). In contrast, the majority of hyx RNAi metaphase NSCs assembled scarce spindle microtubule mass following recovery; at 2 min after recovery, only 14.6% of metaphase NSCs formed mitotic spindles, which were still shorter and thinner than the spindles formed in the control NSCs. MTOCs remained weak or absent in these hyxdepleted NSCs (Fig 4B). Quantification of microtubule intensity suggested that in all of the time points (except for t = 0), microtubule intensity in hyx-depleted NSCs at both interphase and metaphase were significantly reduced compared with the control (Fig 4B and 4D). Taken together, we propose that Hyx plays a central role in the formation of interphase microtubule asters and the mitotic spindle by promoting microtubule growth in NSCs.

Centrosomal localization of Msps, D-TACC, and Polo is dependent on Hyx function in NSCs
Msps is an XMAP215/ch-TOG family protein and a key microtubule polymerase that controls microtubule growth and asymmetric division of NSCs in Drosophila larval central brains [44,74]. In control interphase NSCs, Msps  As Polo, another key centrosomal protein, is critical for the assembly of interphase microtubule asters and asymmetric cell division [27,75], we tested whether Hyx regulates the localization of Polo at the centrosomes. In control interphase NSCs, Polo was strongly detected at the centrosome marked by Asl ( Fig 5A). In contrast, Polo was almost completely absent in 86.8% of hyx HT622 and 58.8% of hyx w12-46 interphase NSCs ( Fig 5A). Furthermore, in control metaphase NSCs, Polo mainly appeared on the centrosomes and kinetochores and weakly on the mitotic spindle ( Fig 5B). However, 80.0% of hyx HT622 and 75% of hyx w12-46 metaphase NSCs lost Polo loci, and the remaining NSCs only had a weak Polo signal (Fig 5B). Similarly, upon hyx knockdown, Polo was almost completely lost from the centrosomes in 84.6% of interphase NSCs and 78.3% of metaphase NSCs (S8A and S8B Fig). The fluorescence intensity of Polo was significantly reduced at the centrosomes in hyx-depleted interphase and metaphase NSCs (Figs 5C and S8C).
Centrosomal protein AurA inhibits NSC overgrowth and regulates centrosome functions by directing the centrosomal localization of D-TACC and Msps [35,76]. We sought to examine whether the centrosomal localization of AurA is dependent on Hyx. AurA is clearly observed at the centrosomes marked by Asl in control interphase and metaphase NSCs (Fig 5D and 5E). Remarkably, AurA was nearly undetectable at the centrosomes in 87.2% of hyx HT622 and 75.0% hyx w12-46 interphase NSCs (Fig 5D). The fluorescence intensity of AurA decreased to  (Fig 5D and 5F). Likewise, AurA levels were significantly reduced in 88.9% of hyx HT622 and 50% of hyx w12-46 metaphase NSCs (Fig 5E and 5F). Furthermore, AurA was diminished in 100% of interphase NSCs and 56.7% of metaphase NSCs upon hyx knockdown (S8D- S8F Fig). Taken together, our data show that Hyx plays an essential role in centrosome assembly and functions by recruiting major centrosomal proteins to the centrosomes in NSCs.

The disruption of NSC polarity and centrosome assembly is a direct consequence of hyx depletion, but not aging
To rule out the possibility that the disruption of NSC polarity and centrosome assembly was due to consequence of aging in late larval stages, we examined NSC polarity proteins and centrosomal proteins at 24 h ALH, a time point when NSCs exit quiescence and reenter the cell cycle [77]. At  In addition, in late larval stages, hyx RNAi hyx HT622/+ showed a stronger NSC overproliferation phenotype than that observed in hyx knockdown alone (Fig 1B and 1C); 84.3% of type I lineages and 93.3% of type II lineages with multiple NSCs were observed in hyx RNAi hyx HT622/+ compared with the control with a single NSC per lineage (S10A Fig). Moreover, Hyx protein was diminished in 89.1% of hyx RNAi hyx HT622/+ NSCs, while it was strongly detected in the control (S10B Fig). Consistent with these observations, strong reduction of γtub and Polo protein levels at the centrosomes was observed in both interphase and metaphase NSCs from hyx RNAi hyx HT622/+ (S10C-S10H Fig).
Taken together, the disruption of NSC polarity and centrosome assembly is a direct consequence of hyx loss of function instead of aging.

Hyx is required for centrosome assembly in S2 cells in vitro
To investigate whether Hyx plays a role in centrosome assembly in nonneuronal cells, we knocked down hyx in S2 cells by dsRNA treatment. We found that a centriolar protein, Ana2, remained localized at the centrosomes in metaphase cells (S8G Fig). This suggests that Hyx is not essential for the localization of centriolar proteins in both S2 cells and NSCs. Next, we examined the localization of other centrosomal proteins in S2 cells. Remarkably, D-TACC intensity was significantly decreased at the centrosomes, upon hyx knockdown, in metaphase S2 cells (Fig 6A and 6B). Consistent with these observations, the intensity of α-tub was also decreased by 0.65-fold on mitotic spindles (Fig 6D and 6E). These in vitro data support our observations in the larval brain and indicate that Hyx regulates microtubule growth and the localization of centrosomal proteins. Polo is undetectable in interphase S2 cells, unlike its significances were determined two-way ANOVA with multiple comparison were performed in C and F. In C, �� p = 0.0044, ns = 0.0634; in F, �� p = 0.0040, � p = 0.0382. Scale bars: 5 μm. The underlying data for this figure can be found in the S1 Data. Asl, Asterless; AurA, Aurora-A; Hyx, Hyrax; MARCM, mosaic analysis with a repressible cell marker; NSC, neural stem cell.
https://doi.org/10.1371/journal.pbio.3001834.g005 robust localization in NSCs during the interphase. Consistent with our in vivo observations, we found that the overall intensity of Polo was significantly reduced to 0.67-fold in the dividing metaphase cells upon hyx knockdown (Fig 6A and 6C). Also, γ-tub intensity at the centrosomes marked by Ana2 was similar to that observed in the control (S8G and S8H Fig). The different observations in S2 cells and larval brains are likely due to incomplete depletion of hyx in S2 cells and/or different underlying mechanisms in vitro.
To further probe how Hyx regulates centrosome assembly, we examined the ultrastructure of Cnn and γ-tub using super-resolution imaging. Cnn and γ-tub formed "doughnut-like" rings surrounding the centriolar protein Asl, at the centrosomes, in 94.9% and 92.9% of control metaphase cells, respectively (Fig 6F and 6G). Remarkably, Cnn and γ-tub failed to form the ring patterns or formed a ring with reduced inner size at the centrosomes in 51.3% and 53.4% of hyx knockdown mitotic cells, respectively (Fig 6F and 6G). These observations suggest that Hyx is required for the proper recruitment of Cnn and γ-tub at the centrosomes in S2 cells.

Hyx directly regulates the expression of polo and aurA in vitro
Next, we investigated whether Hyx directly regulates the expression of polo and aurA, the 2 key centrosomal proteins. We performed chromatin immunoprecipitation (ChIP) coupled with quantitative PCR (ChIP-qPCR) in S2 cells. After normalizing against "Pre-serum" (1-fold), only a 1.37-fold increase was seen for the negative control. In contrast, 2.94-fold enrichment was observed for orb2 promoter, a positive control. Moreover, we found Hyx binds to the promoter region of polo (new Fig 6I; 2.63-fold and 2.95-fold using 2 pairs of primers). Hyx also binds to the promoter region of aurA (Fig 6I; 2.85-fold), but not numb (Fig 6I; 1.64-fold). Therefore, Hyx binds to the promoter region of both polo and aurA.
We performed the luciferase assay to verify the direct binding of Hyx to the polo promoter. The endogenous Hyx in S2 cells was able to induce the luciferase reporter activity under the control of polo-promoter (poloPro) normalized against Renilla luciferase activity, but not with the vector control ( Fig 6J). We attempted to overexpress Hyx in S2 cells to test if it further enhances the luciferase activity under the control of the polo promoter. However, overexpression of Venus-tagged full-length hyx (hyx-FL) resulted in severe cell death (54.3%) detected by active Caspase-3 (S10I and S10J Fig; 11.2% cell death in the control), which precluded us from testing the effect of Hyx overexpression on the transcription of polo in the luciferase assay. Next, we sought to knock down hyx with dsRNA treatment in S2 cells and analyze the relative luciferase activity under the control of the polo promoter. The relative luciferase activity from the ds-hyx treatment group was significantly reduced to 0.5-fold compared with 1-fold from the control group (ds-egfp) (Fig 6K) The relative luciferase activity driven by actin5c promoter induced by ds-hyx treatment and ds-egfp groups was indistinguishable (Fig 6L; 1.0-fold versus 1.2-fold). We conclude that hyx can directly bind to the polo promoter region and promotes its transcription.

The expression of centrosome-related genes depends on Hyx function in larval brains
As Parafibromin/Hyx regulates transcriptional events [78], we wondered whether Hyx was required for the expression of genes that are involved in centrosome assembly. To this end, we sought to perform reverse transcription quantitative real-time PCR (RT-qPCR) to detect differential transcription levels of those genes in larval brains. As both hyx alleles (hyx HT622 and hyx w12-46 ) led to embryonic lethality, we knocked down hyx in the larval brain by RNAi using a ubiquitous driver actin5C-Gal4. However, hyx RNAi under actin5C-Gal4 caused larval lethality after 48 h ALH. Therefore, we performed both RT-qPCR and western blot experiments for hyx RNAi under the control of actin5C-Gal4 at 48 h ALH. hyx mRNA levels were dramatically reduced to 0.23-fold upon hyx knockdown (Fig 6M). The overall Hyx throughout hyx-depleted larval brains was significantly reduced (fluorescence intensity 19.7 ± 8.99 a.u, n = 8 BL), compared with that in control (41.5 ± 13.25, n = 10 BL) (Fig 6N and 6O). Particularly, Hyx levels were dramatically decreased in these hyx-depleted brains (20.4 ± 12.92, n = 99 NSCs), compared with that in control (107.1 ± 29.18, n = 96 NSCs) (Fig 6N and 6O).
Moreover, the western blot result showed that Hyx protein levels were dramatically reduced in hyx RNAi brain under actin5C-Gal4 to 0.34-fold in contrast to 1-fold in control (Fig 6P and  6Q), indicating an efficient hyx knockdown using actin5C-Gal4. Both Polo (0.25-fold) and AurA (0.43-fold) levels normalized against GAPDH levels were dramatically decreased following hyx depletion, compared with 1-fold in control (Fig 6R and 6S).
Taken together, our data suggest that Hyx appears to primarily regulate the expression of polo and aurA in NSCs.

polo and aurA are 2 key downstream targets of Hyx in NSCs
Given that Polo and AurA regulate cell polarity and microtubule functions in NSCs, we sought to investigate whether Polo and AurA are physiologically relevant targets of Hyx in NSCs. We overexpressed polo and aurA in the hyx RNAi knockdown background and found that the ectopic NSC phenotype caused by hyx depletion was significantly suppressed. With the introduction of Venus-polo and aurA into hyx RNAi, the average total NSC number of each brain lobe was significantly reduced to 111.1 ± 10.1 and 114.0 ± 5.1, respectively (Fig 7A and 7B; n = 7 BL and n = 11 BL, respectively) compared with 133.8 ± 4.0 NSCs in hyx RNAi alone ( Fig  7A and 7B; n = 5 BL), close to 98.5 ± 2.8 in control larval brains (Fig 7A and 7B; n = 11 BL), Venus-Polo-(99.2 ± 2.4, n = 10 BL) or AurA-overexpressing brains (98.9 ± 3.6, n = 12 BL). Moreover, the average type I NSC number per lineage in hyx RNAi with Venus-polo and aurA overexpression was decreased to 1.7 ± 0.13 (n = 66 NSC lineage) and 2.3 ± 0.32 (n = 27 NSC lineage) (Fig 7C; mean ± SEM), significantly lower than 3.9 ± 0.77 (n = 27 NSC lineage) in hyx RNAi with β-gal RNAi. Likewise, the average type II NSC number per lineage in hyx RNAi with polo and aurA overexpression was significantly dropped to 4.6 ± 0.40 (n = 36 NSC lineage) and 4.5 ± 0.50 (n = 21 NSC lineage), respectively, compared with 7.7 ± 1.4 (n = 19 NSC lineage) in hyx RNAi with β-gal RNAi. Only 1 NSC per lineage was observed in both type I and type II NSC lineages from β-gal RNAi control, Venus-polo overexpression control, and aurA overexpression control.
Importantly, cell polarity defects caused by hyx depletion were also significantly suppressed by overexpression of polo and aurA in NSCs. The majority (73.4%) of hyx RNAi metaphase NSCs had lost aPKC polarity and the remaining 26.6% of NSCs had a weak aPKC crescent ( Fig  7D and 7E; n = 55). Remarkably, 4.1% of hyx RNAi NSCs with Venus-polo overexpression formed a strong aPKC crescent and 48.9% showed a weak aPKC crescent (Fig 7D and 7E; n = 59 NSC). Similarly, a strong Mira crescent was observed in 1.4% of hyx RNAi NSCs with Venus-polo overexpression and 57.0% formed a weak Mira crescent (Fig 7D and 7F; n = 59), compared with 29.8% of NSCs with a weak Mira crescent in hyx knockdown alone (Fig 7D  and 7F; n = 55). Likewise, aurA overexpression can dramatically restore the asymmetric localization of both aPKC and Mira in hyx RNAi NSCs. Upon aurA overexpression, a strong aPKC crescent was seen in 6.5% of hyx RNAi NSCs and a weak aPKC crescent was formed in 54.7% of NSCs (Fig 7D and 7E; n = 51 NSC). Similarly, upon aurA overexpression, 6.5% of hyx RNAi NSCs formed a strong Mira crescent and 51.5% showed a weak Mira crescent (Fig 7D and 7F; n = 51 NSC). These suppressions were partial because the vast majority of metaphase NSCs from control (n = 58), Venus-Polo-(n = 25), or AurA-overexpression (n = 24) assembled strong aPKC and Mira crescent on the cortex (Fig 7D). Moreover, the shorter spindle phenotype in hyx RNAi was also significantly suppressed by polo and aurA overexpression (Fig 7G  and 7H; control, 0.90 ± 0.05, n = 18 NSCs; 0.82 ± 0.06, n = 23 NSCs; 0.92 ± 0.06, n = 22 NSCs; 0.91 ± 0.06, n = 27 NSCs). These results support our conclusion that both Polo and AurA are physiologically relevant targets of Hyx in NSCs. Therefore, down-regulated polo and aurA expression likely accounts for various defects, such as loss of asymmetry of apical and basal proteins and centrosome/microtubule abnormalities, observed in hyx-depleted NSCs. Polo can be directly activated by AurA in Drosophila mitotic NSCs [79]. Our data support a model in which Hyx promotes the expression of polo and aurA in NSCs and, in turn, regulates cell polarity and centrosome/microtubule assembly (Fig 7I).
Discussion. In this study, we established the essential role of Hyx, the Drosophila ortholog of Parafibromin, during the development of the CNS. We show that Hyx governs NSC asymmetric division and inhibits ectopic NSC formation in the central brains of Drosophila larvae. We also demonstrate that Hyx plays a novel function in the formation of microtubule asters and mitotic spindles in interphase NSCs. Particularly, Hyx is important for the localization of PCM proteins to the centrosomes in dividing NSCs and S2 cells. Therefore, this is the first study to demonstrate that Parafibromin/Hyx plays a critical role in the asymmetric division of NSCs, by maintaining NSC polarity and regulating microtubule/centrosomal assembly in these cells.
It is established that Drosophila Hyx is essential for embryogenesis and wing development [51]. In this study, we provide the first evidence that Hyx is crucial for Drosophila larval brain development. Furthermore, we showed that Hyx is essential for the polarized distribution of proteins in dividing NSCs, indicating a novel role for Hyx in regulating NSC apicobasal polarity. We also found that Hyx is required for the centrosomal localization of AurA and Polo kinases in NSCs, 2 brain tumor suppressor-like proteins that regulate asymmetric cell divisions [26][27][28]. Mechanistically, Hyx promotes the expression of both aurA and polo by directly binding to the promoter region of these 2 genes. Hyx does not seem to affect the expression of a few other polarity genes we have tested, such as apkc, baz, pins, and numb. Overexpression of either polo or aurA partially suppressed ectopic NSC formation and NSC polarity defects caused by hyx depletion, suggesting that polo or aurA are 2 physiological relevant targets of Hyx in NSCs. Therefore, our study identifies a previously unknown link between Hyx and

PLOS BIOLOGY
these cell cycle regulators, raising an interesting possibility that similar regulatory mechanisms may exist in other types of dividing cells, including cancer cells.
Human Parafibromin is a well-known tumor suppressor in parathyroid carcinomas and many other types of cancers [45][46][47][48]. Parafibromin primarily regulates transcriptional events and histone modifications [49,50]. It is also known to inhibit cell proliferation by the blockage of a G1 cyclin, Cyclin D1, and the c-myc proto-oncogene [51,78,80]. We show that Hyx controls polo expression likely through a direct transcriptional regulation. Among Paf1 complex components, only rtf1 RNAi resulted in weak ectopic type II NSCs (but no ectopic type I NSCs). This finding suggests that the function of Hyx in asymmetric cell division of NSCs is largely independent of other components of the Paf1 complex. As our data support the role of Hyx in transcriptional regulation, it appears that Hyx alone without other Paf1 components might be sufficient to promote the target gene transcription. Nevertheless, we cannot rule out the possibility that other Paf1 components are involved in the asymmetric division to a much lesser extent or have redundant functions.
In addition to its role in promoting asymmetric cell divisions and the establishment of apicobasal cell polarity, we provide compelling data that Hyx plays a novel role in regulating microtubule growth and centrosomal assembly in NSCs and S2 cells. Hyx was found to be important for the formation of interphase microtubule asters and the mitotic spindle. We also showed that it is required for the centrosomal localization of major PCM proteins in NSCs, including γ-tub, Cnn, AurA, and Polo. AurA is known to recruit γ-TuRC, Cnn, and D-TACC to the centrosomes [76]. Centrosomal AU : PleasecheckandconfirmthattheeditstothesentenceCentrosom proteins, such as Msps/D-TACC and Cnn, may not be well recruited due to defective centrosome maturation and not a direct effect of Hyx on their expression. Unlike in dividing neuroblasts, in S2 cells, γ-tubulin is important for the nucleation of both centrosomal microtubules and noncentrosomal microtubules, i.e., chromatinmediated microtubule assembly [81]. Therefore, in S2 cells, even with a reduction of Polo and AurA on the centrosomes, γ-tubulin might be recruited to the spindle poles in a centrosomeindependent manner. The new role for Hyx in regulating microtubule growth and asymmetric divisions in NSCs proposed in this study is consistent with our previous finding that NSC polarity is dependent on microtubules [44].
In addition to its predominant localization and functions in the nucleus, Parafibromin is also known to exist in the cytoplasm, where it regulates apoptosis by directly targeting p53 mRNA [82]. Human Parafibromin directly interacts with actin-binding proteins, actinin-2 and actinin-3, during the differentiation of myoblasts [83], suggesting that Parafibromin might regulate the actin cytoskeleton. Interestingly, C. elegans Ctr9 is required for the microtubule organization in epithelial cells during the morphogenesis of the embryo [84]. Therefore, the function of Hyx/Parafibromin in regulating centrosomal assembly is likely a general paradigm in cell division regulation, which might be disrupted in cancer cells. Parafibromin/ HRPT-2 is expressed in both mouse and human brains [52]. Deletion of Hrpt2 in mouse embryos results in early lethality and a developmental defect of the brain, suggesting that Parafibromin may play a role in CNS development [85]. Further investigations on the likely conserved functions of mammalian Parafibormin in NSC divisions and microtubule growth are warranted in future studies.
All experiments were carried out at 25˚C, except for RNAi knockdown or overexpression experiments that were performed at 29˚C.

Immunohistochemistry
Third instar Drosophila larvae were dissected in PBS, and larval brains were fixed in 4% EMgrade formaldehyde in PBT (PBS + 0.3% Triton-100) for 22 min. The samples were processed for immunostaining as previously described [77]. For α-tubulin immunohistochemistry, larvae were dissected in Shield and Sang M3 medium (Sigma-Aldrich), supplemented with 10% FBS, followed by fixation in 10% formaldehyde in Testis buffer (183 mM KCl, 47 mM NaCl, 10 mM Tris-HCl, and 1 mM EDTA (pH 6.8)), supplemented with 0.01% Triton X-100. The fixed brains were washed once in PBS and twice in 0.1% Triton X-100 in PBS. Images were taken with an AxioCam HR camera (with 1.5× to 8× of digital zoom) of a LSM710 confocal microscope system (Axio Observer Z1; ZEISS), using a Plan-Apochromat 40×/1.3 NA oil differential interference contrast objective. The brightness and contrast of the images obtained were adjusted using Adobe Photoshop or Fiji (imageJ).
The primary antibodies used were the following: rabbit affinity-purified anti-Hyx/Cdc73

Spinning disc super-resolution imaging
Super-resolution Spinning Disc Confocal-Structured Illumination Microscopy (SDC-SIM) was performed on a spinning disk system (Gataca Systems) based on an inverted microscope (Nikon Ti2-E; Nikon) equipped with a confocal spinning head (CSU-W; Yokogawa), a Plan-Apo objective (100×1.45-NA), and a back-illuminated sCMOS camera (Prime95B; Teledyne Photometrics). A super-resolution module (Live-SR; GATACA Systems) based on structured illumination with optical reassignment technique and online processing leading to a 2-time resolution improvement [90] is included. The maximum resolution is 128 nm with a pixel size of 64 nm in super-resolution mode. Excitation light at 488 nm/150 mW (Vortran) (for GFP), 561 nm/100 mW (Coherent) (for mCherry/mRFP/tagRFP) and 639 nm/150 mW (Vortran) (for iRFP) was provided by a laser combiner (iLAS system; GATACA Systems), and all image acquisition and processing were controlled by the MetaMorph (Molecular Device) software. Images were further processed with imageJ.

Clonal analysis
MARCM clones were generated as previously described [54]. Briefly, larvae were heat shocked at 37˚C for 90 min at 24 h ALH and 10 to 16 h after the first heat shock. Larvae were further aged for 3 d at 25˚C, and larval brains were dissected and processed for immunohistochemistry. To generate type II NSC clones, UAS lines were crossed to the type II driver (worniu (wor)-Gal4, ase-Gal80 ts ; UAS-CD8-GFP) at 25˚C and shifted to 29˚C at 24 h ALH. Wandering third instar larvae were dissected after incubation for 3 or 4 d at 29˚C. Z-stacks were acquired and NSC number per clone/lineage was manually counted. Percentage of ectopic NSCs refers to the percentage of NSC clones or lineages with ectopic NSCs out of total number of clones/ lineages scored in this study. All clones in the larval brains scored in this study were NSC clones.

Time-lapse recording
The time-lapse recording was performed as described [44]. The whole-mount brain expressing G147-GFP was used to analyze the asymmetric cell division of NSCs. The brain was dissected and loaded into a Lab-Tek chambered coverglass (Thermo Fisher Scientific) filled with dissecting medium that is supplemented with 2.5% methyl cellulose (Sigma-Aldrich). The time-lapse images of NSC divisions were acquired every 30 s on a confocal microscope (LSM 710; ZEISS). The video was processed with ImageJ and displayed at 15 frames per second.

Microtubule regrowth assay
The microtubule regrowth assay was performed as described previously [44]. Third instar larval brains were dissected in Shield and Sang M3 insect medium (Sigma-Aldrich) supplemented with 10% FBS, and microtubules were depolymerized by incubating the larval brains on ice for 40 min. The brains were allowed to recover at 25˚C for various time periods to facilitate microtubule regrowth. The brains were immediately fixed in 10% formaldehyde in testis buffer (183 mM KCl, 47 mM NaCl, 10 mM Tris-HCl, and 1 mM EDTA (pH 6.8)) supplemented with 0.01% Triton X-100. The fixed brains were washed once in PBS and twice in 0.1% Triton X-100 in PBS, following which they were processed for immunohistochemistry. The mean immunofluorescence intensity of α-tub detected on astral microtubules proximal to MTOC at interphase and spindle microtubules at metaphase were quantified on selected regions of the same size with ImageJ.

S2 cell culture, transfection, and quantitative RT-PCR
Cell culture. Drosophila S2 cells were cultured in Express Five SFM (Thermo Fisher Scientific), supplemented with 2 mM glutamine (Thermo Fisher Scientific), at 25˚C.
Double-stranded RNA (dsRNA) production and interference. DNA fragments, approximately 470 bp in length for ds-egfp as control and 825 bp in length for ds-hyx, were amplified using PCR. Each primer used in the PCR contained a 5 0 T7 RNA polymerase binding site (TAATACGACTCACTATAGGG) followed by sequences specific for the targeted genes. The PCR products were purified by using the QIAquick PCR Purification Kit (Cat No. 28106). The purified PCR products were used as templates for the synthesis of dsRNA, by using a MEGA-SCRIPT T7 transcription kit (Ambion, Austin, TX). The dsRNA products were ethanol precipitated and resuspended in water. The dsRNAs were annealed by incubation at 65˚C for 30 min followed by slow cooling to room temperature. To ensure that the majority of the dsRNA existed as a single band, 1 μg of dsRNA was analyzed by 1% agarose gel electrophoresis. S2 cells were cultured in 24-well plates at 50% to 90% confluency in 500 μl of medium. Cells were treated with 5 μg dsRNA and collected 72 h after transfection for mRNA extraction and immunostaining.
The primers used for dsRNA synthesis were the following: ds-egfp-forward:

Chromatin immunoprecipitation
ChIP was performed according to the manufacturer's protocol (Cell Signaling, #9005). Sonicated lysates were used for ChIP with antibodies against Hyx (final bleed, J. T. Lis) and preimmune serum as a control. Immunoprecipitated DNA was analyzed by quantitative real-time PCR using specific primers to the potential promoter region of various centrosomal genes and negative control-the intergenic sequence at 5 kb downstream of numb genome as well as the positive control targeting the potential orb2 promoter region:

Quantification and statistical analysis
Drosophila larval brains from various genotypes were placed dorsal side up on confocal slides. The confocal z-stacks were taken from the surface to the deep layers of the larval brains. For each genotype, at least 10 NSCs were imaged and ImageJ or Zen software was used for quantifications. The localization of polarity proteins was scored by 3 categories: "Strong crescent," "Weak crescent," and "No crescent." Statistical analysis was essentially performed using GraphPad Prism 9. Unpaired two-tail t tests were used for comparison of 2 sample groups, and one-way ANOVA or two-way ANOVA followed by Sidak's multiple comparisons test was used for comparison of more than 2 sample groups. All data are shown as the mean ± SD. Statistically nonsignificant (ns) denotes p > 0.05, � denotes p <0.05, �� denotes p <0.01, ��� denotes p <0.001, and ���� denotes p < 0.0001. All experiments were performed with a minimum of 2 repeats. In general, n refers to the number of NSCs counted unless otherwise indicated.  Hyx is dramatically decreased upon hyx knockdown under the control of actin5C-Gal4. Western blotting analysis of larval brain protein extracts of control (UAS-β-Gal RNAi; UAS-β-Gal RNAi) and hyx knockdown with UAS-Dicer2 (hyx RNAi; UAS-Dicer2 RNAi) driven by actin5C-Gal4 at 48 h ALH. Blots were probed with anti-Hyx antibody (upper panels) and anti-GAPDH antibody (lower panels). Both "Unlabeled" (left panels) and "Labeled" (right panels) uncropped original blots were provided. A protein ladder was indicated on the left of the "Labeled" membrane. Cropped images used in Fig 6P were boxed and antibodies used were indicated by arrows. Original uncropped western blot for Fig 6R. Polo and AurA protein levels were significantly decreased upon hyx knockdown under the control of actin5C-Gal4. Western blotting analysis of 48 h ALH larval brain extracts of control (UAS-β-Gal RNAi; UAS-β-Gal RNAi) and hyx knockdown with UAS-Dicer2 (hyx RNAi; UAS-Dicer2 RNAi) under the control of actin5C-Gal4. Blots were probed with anti-Polo antibody (upper panels), anti-AurA antibody (middle panels), and anti-GAPDH antibody (lower panels). Both "Unlabeled" (left panels) and "Labeled" (right panels) uncropped original blots were provided. A protein ladder was indicated on the left of the "Labeled" membrane. Cropped images used in Fig 6R were boxed and antibodies used were indicated by arrows. (PDF)