Drosophila septin interacting protein 1 regulates neurogenesis in the early developing larval brain

Neurogenesis in the Drosophila central brain progresses dynamically in order to generate appropriate numbers of neurons during different stages of development. Thus, a central challenge in neurobiology is to reveal the molecular and genetic mechanisms of neurogenesis timing. Here, we found that neurogenesis is significantly impaired when a novel mutation, Nuwa, is induced at early but not late larval stages. Intriguingly, when the Nuwa mutation is induced in neuroblasts of olfactory projection neurons (PNs) at the embryonic stage, embryonic-born PNs are generated, but larval-born PNs of the same origin fail to be produced. Through molecular characterization and transgenic rescue experiments, we determined that Nuwa is a loss-of-function mutation in Drosophila septin interacting protein 1 (sip1). Furthermore, we found that SIP1 expression is enriched in neuroblasts, and RNAi knockdown of sip1 using a neuroblast driver results in formation of small and aberrant brains. Finally, full-length SIP1 protein and truncated SIP1 proteins lacking either the N- or C-terminus display different subcellular localization patterns, and only full-length SIP1 can rescue the Nuwa-associated neurogenesis defect. Taken together, these results suggest that SIP1 acts as a crucial factor for specific neurogenesis programs in the early developing larval brain.


Ensembles of neurons are produced by a limited number of neural stem cells (called neuroblasts in Drosophila)
and assemble into complex neural circuits. These circuits comprise the functional nervous system required for animal survival and reproduction. The Drosophila nervous system is a well-characterized model that is widely used for investigating the molecular and genetic programs crucial for neurogenesis. In the adult Drosophila central brain, each hemisphere is composed of neurons from around 100 neuroblast-derived lineages and has a total number of roughly 11,000 neurons [1][2][3] . Notably, a previous study showed that the final neuronal composition (i.e., cell numbers and subtypes of neurons) for two olfactory neural lineages-anterodorsal projection neurons (adPNs) and lateral antennal lobe neurons-is extremely difficult to perturb, even under the harsh challenge of dietary protein starvation 4 . In contrast, the neurogenesis of certain groups of neurons like Kenyon cells, the intrinsic neurons of the learning and memory center (the mushroom body), is highly plastic and uncoupled from organismal growth and development 4 . Tight regulation, robustness and plasticity are key characteristics of neurogenesis that enable an organism to produce a functional nervous system. Therefore, revealing the molecular and genetic mechanisms that ensure generation of appropriate neuronal numbers and subtypes is an important but challenging task for neurobiologists.
Neurogenesis of the Drosophila central brain occurs during two major developmental periods 5 . At the embryonic stage, neuroblasts divide 10-20 times to produce the neurons that construct larval-specific neural circuits. Then, at postembryonic stages, the neuroblasts undergo 100-200 rounds of the cell cycle to produce more neurons; these postembryonic neurons are assembled with some of the embryonic-born neurons to constitute adult-specific neural circuits. Between these two waves of neurogenesis, neuroblasts undergo cell cycle arrest (at the end of embryogenesis) and resume proliferation at the early larval stage 5 . Interestingly, the cell bodies of quiescent neuroblasts become enlarged prior to their awakening and re-entry into the cell cycle; this reactivation process is regulated by Hippo, insulin receptor, and target of rapamycin signaling pathways at the early larval stage [6][7][8] . After reactivation of quiescent neuroblasts, the rates of neurogenesis for many neural lineages are accelerated, as evidenced by increased EdU/BrdU incorporation in the brain at early to mid-larval stages 9,10 . Most proliferative neuroblasts eventually lose their ability to divide at the early pupal stage, likely due to a switch in energy metabolism induced by the steroid hormone, ecdysone, and the Mediator complex. This switch results in the shrinkage of neuroblast cell body size and gradually leads neuroblasts to the exit the cell cycle 11 . Since neurogenesis is a protracted process with dynamic rates of neuron production, it is probable that OPEN Institute of Cellular and Organismic Biology, Academia Sinica, Taipei, Taiwan. * email: samhhyu@gate.sinica.edu.tw specific molecular and genetic programs govern neurogenesis in a developmental stage-dependent manner, e.g., differential activities at early versus late larval stages.
In our ongoing MARCM (mosaic analysis with a repressible cell marker) 12 -based genetic screen for modulators of neurogenesis, we identified a novel mutation, Nuwa. In this mutant line, we found impaired neurogenesis in all examined neural lineages when the homozygous mutation was induced at the early larval but not embryonic or late larval stages. Interestingly, the gene defect responsible for the Nuwa-associated neurogenesis phenotype was mapped to Drosophila septin interacting protein 1 (sip1) 13 . During development, SIP1 expression was enriched in neuroblasts and RNAi knockdown of sip1 using a neuroblast driver could recapitulate Nuwa-associated aberrant brain features, including smaller brain size. Finally, we found that full-length SIP1 protein and truncated SIP1 proteins displayed preferential subcellular localizations, and only full-length SIP1 protein could rescue the Nuwa-associated neurogenesis defects. Taken together, these results suggest that SIP1 acts as a crucial factor for neurogenesis processes in the early developing larval brain.

Results
Induction of the P 111477 mutation at the early larval stage significantly impairs vPN neurogenesis. We conducted a MARCM-based screen using GAL4-MZ699, which labels most ventral olfactory projection neurons (vPNs). With this ongoing screen we seek to identify mutations that affect vPN morphologies and hopefully provide clues about the genetic and molecular mechanisms underlying Drosophila central brain development (Fig. 1a). In our MARCM experiments, the labeling of GAL4-MZ699-positive vPNs is dependent on the induction of FRT (flippase recognition target)-mediated mitotic recombination in the vPN neuroblasts. Therefore, a gradual reduction in the number of labeled vPNs is expected when MARCM neuroblast clones are induced from early to late developmental stages (Fig. 1b). In agreement with this expectation, cell numbers of vPNs in wild-type flies were counted as 64. 8  Neurogenesis is compromised in various brain regions when the Nuwa mutation is induced at early but not late larval stage. Based on DGRC annotations, P 111477 carries two P-element insertions, l (2) k07109a and l(2)k07109b (referred to as P 07109a and P 07109b , respectively) on the FRT 40A background, which allows for MARCM-related experiments. P 07109a is inserted into an unknown gene at cytolocation 25F2, whereas P 07109b is inserted in the Fas3 gene at cytolocation 36F2. To investigate whether loss of Fas3 function is responsible for the neuronal production defect, we examined the vPN number in an independent Fas3 mutant (DGRC 111,717, referred to as P 111717 ). However, we found a normal number of vPNs arose from the P 111717 mutant neuroblast clones, suggesting that the absence of Fas3 alone did not compromise the production of vPNs (Supplemental Fig. 2a). In addition, we utilized a P-element insertion line (DGRC 102,523, referred to as P 102523 ), which is annotated as a single P 07109a insertion without the P 07109b insertion, to test whether the vPN neurogenesis defect is caused by the P 07109a insertion. We assembled P 102523 into the FRT 40A background, and as expected, the production of vPNs was impaired in the mutant. As such, the mutant had significantly fewer vPNs (P 102523 mutant samples: 3.8 ± 1.1, n = 5, P < 0.01; Supplemental Fig. 2b,d) than wild-type samples when MARCM neuroblast clones were induced at NHL-24 h ALH. In contrast, vPN neurogenesis was relatively normal, and similar vPN numbers were found in wild-type and P 102523 mutant samples (24 ± 4.4, n = 3, P > 0.1; Supplemental Fig. 2c,d) when MARCM neuroblast clones were induced at 72 h ALH. Since these results together suggested that P 07109a is probably the mutation in P 111477 that compromises vPN neurogenesis, the P 102523 mutation was used in most of the subsequent experiments.
We wondered whether the neurogenesis defect caused by the P 102523 mutation is restricted to vPNs or if it broadly occurs in other groups of neurons. Therefore, we examined neurons generated in various brain regions (Fig. 2a), including neurons in the subesophageal zone (SEZs; Fig. 2b-e), ventral olfactory interneurons in the AL (vLNs; Fig. 2f-i) and neurons in the ventrolateral protocerebrum (VLPs; Fig. 2j-m). Notably, all of the examined neurons displayed severe neurogenesis defects in the P 102523 mutation when MARCM neuroblast clones were induced at NHL-24 h ALH (Fig. 2b,d,f,h,j,l). On the other hand, we did not observe obvious neurogenesis defects among any examined groups of neurons in the P 102523 mutation when MARCM neuroblast clones were induced at 72 h ALH (Fig. 2c,e,g,i,k,m). Of note, the neuron morphologies were generally aberrant in P 102523 mutant clones for all neuronal groups examined in Figs. 1  www.nature.com/scientificreports/ thus far suggested that the gene disrupted by the P 102523 insertion is generally required for neurogenesis in various brain regions at the early but not late larval stages. This function of the unknown gene reminded us of the legend of an ancient goddess, Nuwa, who is considered to be the creator of mankind in Chinese myths. Therefore, we named the mutation Nuwa to reflect its essential role in the control of neurogenesis at the early larval stage.
Embryonic-born adPNs are produced normally when the Nuwa mutation is induced at the embryonic stage but postembryonic-born adPNs are not. Since neurogenesis in the Drosophila central brain occurs at both embryonic and postembryonic stages, we also wondered whether Nuwa dictates neurogenesis at times other than the early larval stage. In particular, we wondered whether Nuwa is also required for neurogenesis at the embryonic stage. Since adPNs are well-characterized in terms of neuronal numbers and subtypes produced at both embryonic and postembryonic stages, we focused on adPNs to assess the requirement of Nuwa at the embryonic stage 14 . First, we confirmed that a neurogenesis defect was indeed present in adPNs when the Nuwa mutation was induced at the early larval stage (Supplemental Fig. 3). We then conducted twin-spot MARCM experiments using GAL4-GH146, which labels 15 types (with a total cell number of 15) of embryonic-born adPNs as well as the first 12 types (with a total cell number of around 32) of larval-born adPNs, to comprehensively analyze the Nuwa-associated neurogenesis defect in adPNs 14 (Fig. 3a). In wild-type  www.nature.com/scientificreports/ animals, a VM3a adPN (an embryonic-born adPN), which was associated with around 35 adPNs (both embryonic-and larval-born adPNs), was labeled when a twin-spot MARCM clone was induced at the embryonic stage (Fig. 3b). In contrast, a VM3a wild-type adPN associated with three Nuwa mutant adPNs was seen in a twin-spot MARCM clone when the twin-spot MARCM clone was induced at a similar embryonic stage (Fig. 3c). Notably, the dendrites of these three green Nuwa mutant adPNs were arborized in DM3, VM3 and DL4 glomeruli of the AL (Fig. 3c). This arborization pattern implied that the cells were the last three types of embryonic-born adPNs and further suggested that no larval-born adPN was generated when Nuwa was mutated in adPNs at the embryonic stage. Moreover, additional analyses of twin-spot MARCM clones derived from the Nuwa mutant adPN neuroblast all led to a similar conclusion, i.e., the generation of embryonic-born adPNs was unaffected, but production of larval-born adPNs was prevented when adPN neuroblasts became Nuwa mutants at the embryonic stage (Supplemental Fig. 4). Collectively, our data showed that neurogenesis of embryonic-born, but not larvalborn, adPNs was relatively normal when the Nuwa mutation was induced starting from the embryonic stage in twin-spot MARCM experiments.

Molecular characterization and identification of the Nuwa mutation as a loss-of-function in Drosophila septin interacting protein 1.
To identify the gene that is interrupted by the insertion of P 07109a and causes the Nuwa-associated neurogenesis phenotype, we used an inverse PCR method 15 to identify the genomic DNA fragments flanking P 07109a (Fig. 4a). However, we encountered two unexpected results after recovering P-element-flanking genomic DNA fragments from P 102523 , P 111477 , FRT 40A and P 102523 , FRT 40A mutant lines. First, the P 07109b insertion still appeared to be present in the original P 102523 mutant line, since a genomic DNA fragment flanking the P-element was matched to Fas3 (Fig. 4b,c; Supplemental Fig. 5). In contrast, the P 07109b insertion was not identified in P 111477 , FRT 40A or P 102523 , FRT 40A mutant lines, since we did not recover P-element-flanking DNA fragments for Fas3 from these lines ( Fig. 4c; Supplemental Fig. 5). In retrospect, it probably should not have been overly surprising that P 07109b could be lost in the process of performing genetic crosses to generate P 111477 , FRT 40A and P 102523 , FRT 40A flies because P 07109b is located between FRT 40A and P 07109a . The second unexpected result was that P 07109a in P 111477 , FRT 40A and P 102523 , FRT 40A mutant lines consisted of P-elements inserted into two different genes, Dpr-interacting protein θ (DIP-θ; P 07109a-1 is inserted 76 bp downstream of the beginning of the DIP-θ transcript at cytolocation 25F2) and CG11030 (P 07109a-2 is inserted 21 bp downstream of the beginning of the CG11030 transcript at cytolocation 25F4) (Fig. 4a-c). Based on this information, we further investigated whether loss of CG11030 or DIP-θ function is responsible for the Nuwa-associated neurogenesis defect. We overexpressed cDNA transgenes for CG11030 or DIP-θ in Nuwa mutant MARCM neuroblast clones using a pan-cell driver (Act-FRT < stop < FRT-GAL4) 3 and looked for rescue of the neurogenesis defect. However, neither CG11030 nor DIP-θ was able to rescue the Nuwa-associated neurogenesis defect ( Fig. 5a-d), suggesting that loss of CG11030 or DIP-θ function alone may not cause the defect. In addition, these results raised a concern as to whether the Nuwa-associated neurogenesis defect could be derived from a background mutation in chromosome 2L, outside the genomic region of cytolocation 25F2-4, since the Nuwa phenotype was found in MARCM experiments using FRT 40A .
To resolve this issue, we employed two approaches using deficiency and transgenic lines carrying bacterial artificial chromosome (BAC) genomic DNAs to map the genomic region of Nuwa. First, we selected two deficiency lines with deletions of the genomic region at cytolocation 25F1-4, including Df(2L)Exel8016 (genomic region from the DIP-η gene to the DIP-θ gene is deleted) and Df(2L)ED270 (genomic region from the Hsp60c gene to the CG9171 gene is deleted) (Fig. 4a). Since the homozygous mutation of P 102523 insertion is lethal (Supplemental table 2), we performed a complementation test to examine whether animals can survive when carrying trans-heterozygous mutations of P 102523 with Df(2L)Exel8016 or Df(2L)ED270. Interestingly, we found that transheterozygous mutations of P 102523 with Df(2L)ED270, but not Df(2L)Exel8016, were lethal (Supplemental table 2). Thus, we concluded that the putative Nuwa mutation can be found within the genomic region deleted in Df(2L) ED270 but not Df(2L)Exel8016. We then recombined Df(2L)ED270 to the FRT 40A background and conducted MARCM experiments to examine whether Df(2L)ED270 affects neurogenesis like the P 102523 mutation. Indeed, we found that the Df(2L)ED270 mutant neuroblast clones displayed neurogenesis defects similar to those in Nuwa mutants. Substantially fewer vPNs were produced when MARCM neuroblast clones were induced at NHL-24 h ALH, and relatively normal numbers were observed when induction was at 72 h ALH (Supplemental Fig. 6). These results strongly suggested that the Nuwa mutation resides in the genomic region of cytolocation 25F2-4.
In addition to the deficiency line experiments, we also made two customized transgenic lines carrying Pacman BAC genomic DNAs 16 , CH321-13P21 (containing the genomic region of cytolocation 25F1-3, from the DIP-θ gene to the CG12511 gene) and CH321-86B19 (containing the genomic region of cytolocation 25F3-4, from the CG7236 gene to the CG9171 gene), to conduct rescue experiments with the Nuwa mutants (Fig. 4a).

Figure 2.
Neurogenesis is generally compromised in various brain regions when the Nuwa mutation is induced at the early but not late larval stages. (a) Three groups of neurons were used to examine the developmental stagedependent requirement for Nuwa, including neurons in the subesophageal zone (SEZs; panels b-e), ventral olfactory interneurons in the AL (vLNs; panels f-i) and neurons in the ventrolateral protocerebrum (VLPs; panels J-M). (b-m) All three groups of neurons displayed severe neurogenesis defects with the P 102523 (Nuwa) mutation when MARCM clones were induced at NHL-24 h ALH (panels b, d, f, h, j, l), whereas no obvious neurogenesis defects were observed in neurons with the Nuwa mutation when MARCM clones were induced at 72 h ALH (panels c, e, g, i, k, m). Neuropils were revealed by the Brp staining (blue), and background neurons are indicated by arrowheads. Scale bar: 10 μm. www.nature.com/scientificreports/ Intriguingly, the Nuwa-associated neurogenesis defect was rescued by the transgenic line carrying CH321-86B19, but not CH321-13P21, as a significantly higher number of neurons was generated by the Nuwa mutant neuroblast clones (Fig. 5e,f). Taken together, the experimental results from both deficiency lines and genomic transgenic rescue lines indicated that the Nuwa gene resides in the genomic region carried by CH321-86B19 and deleted in the Df(2L)ED270 line. Both of these conclusions rule out the possibility that Nuwa is a background mutation outside the genomic region of cytolocation 25F2-4. Based on these deficiency and genomic DNA rescue results, we further generated additional cDNA transgenic lines carrying individual genes within CH321-86B19 to identify the specific gene involved in the Nuwa-associated neurogenesis defect. Four out of 10 genes, CG7236, CG11147, CG11149 and septin interacting protein 1 (sip1) were initially examined for their abilities to rescue the Nuwa-associated neurogenesis defect (Figs. 4a and 5g,j). Three of the genes, CG7236, CG11147 and CG11149, failed to rescue the Nuwa-associated neurogenesis defect when overexpressed in the Nuwa mutant neuroblast clones (Fig. 5g-i). On the other hand, overexpression of sip1 significantly rescued the Nuwa-associated defect, as a substantial number of neurons were restored in Nuwa mutant neuroblast clones (Fig. 5j). To further test whether the loss of sip1 function indeed causes the Nuwa-associated neurogenesis phenotype, we generated an insertion line, sip1 GCFC . To create this line, we used CRISPR-Cas9 17,18 to insert a DNA fragment that replaces part of the coding region of the sip1 gene (Fig. 6a). As expected, we observed similar vPN neurogenesis defects in the sip1 GCFC mutation when MARCM neuroblast clones were induced at NHL-24 h ALH and 72 h ALH, respectively (Fig. 5k,l). Taken together, these results strongly suggested that the neurogenesis defect seen in the Nuwa mutation was due to the loss of sip1 function. Examples of wild-type and Nuwa mutant twin-spot MARCM clones induced at the embryonic stage. In the wild-type sample (panel b), a VM3a adPN (an embryonic-born adPN in magenta) was associated with around 35 adPNs (containing both embryonic-and larval-born adPNs in green); in the P 102523 (Nuwa) mutant (panel c), a VM3a wild-type adPN (magenta) was associated with three Nuwa mutant adPNs (green, marked by asterisks). Dendrites of the three green Nuwa mutant adPNs were only observed in DM3, VM3 and DL4 [but not DL1 (indicated by dashed arrow)] glomeruli of the AL, indicating that they belong to the last three types of embryonic-born adPNs and further suggesting that no larval-born adPNs were generated. Neuropils were revealed by Brp staining (blue), and background neurons are indicated by arrowheads. Scale bar: 10 μm. www.nature.com/scientificreports/

SIP1 expression is enriched in neuroblasts and sip1
RNAi knockdown using a neuroblast driver causes brain defects. According to the predicted protein sequence and functional domain analysis, SIP1 contains 839 amino acids and at least three domains 13 (Fig. 6a). The first domain is a Tuftelin-interacting protein N-terminal (TIP-N) domain at the N-terminus of SIP1, which has been shown to participate in enamel assembly by interacting with an enamel matrix protein, Tuftelin, in mice 19 . The next domain, called the G-patch domain, is enriched with highly conserved glycines; this type of domain has been found in a number of RNA binding proteins and has a putative function in RNA-related biological processes 19 . The third domain contains a sequence similar to GC-rich sequence DNA-binding factor, transcriptional repressor and histone-interacting proteins (GCFC), and it is presumably involved in transcriptional regulation 13,19 . To detect the distribution of SIP1 protein in the brain, we took advantage of a GFP-tagged transgenic line, sip1::sfGFP (obtained from Vienna Drosophila Resource Center, VDRC318488), generated from a genome-wide forsmid library containing tagged genes with mostly intact regulatory fragments 20 (Fig. 6a). Interestingly, we found that the SIP1::sfGFP expression was enriched in neuroblasts during development (Fig. 6b-d; the relative expression level of SIP1::sfGFP in neuroblasts compared to that in neurons was estimated in Supplemental Fig. 7). Consistent with the SIP1::sfGFP expression pattern, RNAi knockdown of sip1 using a neuroblast driver, worniu-GAL4, resulted in brain defects, including overall smaller brain size, abnormal neuropil architectures and a reduction of neuronal number (Fig. 6e,f; Supplemental Fig. 8a,b). In contrast, no obvious defects were observed when sip1 expression was silenced using a differentiated neuronal driver, synaptobrevin-GAL4 (Fig. 6g). Taken together, these results suggested that SIP1 indeed plays an important role in neurogenesis, and its function may be associated with neuroblasts at the larval stage.

Truncated and full length SIP1 proteins show different subcellular localizations and only full-length SIP1 rescues the Nuwa-associated neurogenesis defect.
To visualize the subcellular localization of SIP1 and gain insight into its functional domains, we generated three transgenic animals that  www.nature.com/scientificreports/ expressed SIP1::GFP fusion proteins with different truncations (Fig. 6a). The full-length SIP1::GFP was constructed by fusing GFP to the C-terminus of full-length SIP1 (containing 839 amino acids), whereas SIP1ΔC::GFP and SIP1ΔN::GFP were made by fusing GFP to the C-terminus of two truncated SIP1 proteins carrying amino acids 1-334 and 333-839 of SIP1, respectively (Fig. 6a). Interestingly, all three SIP1::GFP fusion proteins displayed different subcellular localizations (Supplemental Fig. 9). For instance, the full length SIP1::GFP fusion protein and SIP1ΔN::GFP were observed throughout the entire neuron, with preferential localization of the full length SIP1::GFP at the plasma membrane and in the cytosol (Supplemental Fig. 9a-f). In contrast, the SIP1ΔC::GFP fusion proteins were mostly found in the nucleus (Supplemental Fig. 9g-i). We then asked, do any of these three SIP1::GFP fusion proteins retain the function of SIP1 in regulating neurogenesis? As expected, overexpression of the full-length SIP1::GFP fusion protein could rescue the Nuwa-associated neurogenesis defect (Fig. 6h). However, neither SIP1ΔN::GFP nor SIP1ΔC::GFP was capable of rescuing the Nuwa-associated neurogenesis phenotype (Fig. 6i,j). Since SIP1 was originally identified as a binding protein of Peanut (Pnut, a Drosophila Septin essential for cytokinesis) 21,22 and only the preferentially plasma membrane/cytosol-localized full-length SIP1::GFP could rescue the sip1-deficient neurogenesis phenotype (Fig. 6h), we further examined the possibility that sip1 could affect mitosis. We therefore generated MARCM clones at NHL and analyzed them at 30 h ALH. Interestingly, we found that the average cell numbers of neuroblast clones were 7.08 ± 3.66 and 3.31 ± 1.44 in wild-type (n = 13) and sip1 mutants (n = 16, P < 0.01), respectively (Supplemental Fig. 10. Moreover, neuroblasts containing a mitosis marker, phospho-Histone H3 (H3-P) were 69% and 25% in these wild-type and sip1 MARCM clones, respectively. Taken together, these results suggested that both the N-and C-terminal domains of SIP1 protein are essential for SIP1 function in the regulation of neurogenesis. These domains potentially target SIP1 to different subcellular localizations and are possibly important for mitosis. www.nature.com/scientificreports/

Discussion
Neurogenesis is robustly, tightly and plastically regulated by developmental stage-dependent molecular and genetic programs to produce appropriate neuronal populations in the Drosophila central brain [4][5][6][7][8][9][10][11] . In this study, our MARCM-based genetic screen revealed a novel mutation, Nuwa (loss-of-function in Drosophila sip1), which impaired neurogenesis in all examined neural lineages at the early larval stage but not embryonic or late larval stages (Figs. 1, 2, 3), suggesting a developmental stage-dependent effect of sip1 in Drosophila central brain neurogenesis. Intriguingly, when the homozygous Nuwa mutation was induced at the embryonic stage, embryonicborn adPNs were produced as normal, but larval-born adPNs derived from the same adPN neuroblast were completely absent ( Fig. 3 and Supplemental Fig. 4). Together with the results of our SIP1 expression and RNAi knockdown experiments (Fig. 5), this result implies that the function of sip1 might somehow be linked to the neurogenesis process just after quiescent adPN neuroblast reactivation. However, it remains unclear whether sip1 is broadly required in various neural lineages (other than adPNs) at this time-point. It is also unclear how sip1 affects neurogenesis in the early larval stage, especially during the critical period after quiescent neuroblast reactivation. Future studies involving induction of the Nuwa mutation in other neuroblasts at the embryonic stage and identifying the interacting partners of SIP1 might address this issue. Previously, SIP1 was identified as a binding partner for a Drosophila Septin, Pnut, through a yeast-two hybrid screen 21 . Interestingly, like other Septin proteins, Pnut was shown to be localized to the cleavage furrow of dividing cells during cytokinesis, and pnut loss-of-function mutations resulted in clusters of large, multi-nucleated  Fig. 7. The plasma membrane (outer) and the nuclear membrane (inner) of neuroblasts are indicated by arrows and arrowheads, respectively (labeled in magenta by mCD8:: RFP driven by worniu-GAL4 (wor)). (e-g) RNAi knockdown of sip1 using a GAL4 line expressed in neuroblasts, worniu-GAL4 (wor), but not using a pan-neuronal GAL4 line, synaptobrevin-GAL4 (syb), resulted in aberrant brain morphology, including overall smaller brain size, abnormal neuropil architectures (the AL was indicated by arrows). Excitatory and inhibitory neurons were visualized by choline acetyltransferase (Chat; green) and γ-aminobutyric acid (GABA; magenta) staining. A reduction of GABA-positive neuronal number dorsolateral to the AL was observed in wor > sip1 RNAi knockdown samples (dashed circles; wild-type: 123.5 ± 13.3, n = 6; wor > sip1 RNAi : 39.3 ± 11.3, n = 6; syb > sip1 RNAi : 127.3 ± 24.3, n = 6). (h-j) The Nuwa-associated neurogenesis defect in the P 102523 mutation was rescued by the full-length sip1::GFP transgene, but not sip1ΔN::GFP or sip1ΔC::GFP transgenes. Neuropils were revealed by Brp staining (blue). Scale bar: 10 μm. www.nature.com/scientificreports/ cells, possibly due to a failure of cytokinesis 22 . It is well established that Septin filamentous structures formed by a network of multiple Septin proteins (Pnut included) are essential for cytokinesis in dividing cells [22][23][24] . Although the physical and physiological interactions between SIP1 and Pnut have not yet been demonstrated (other than the yeast two-hybrid result), a hint that SIP1 is potentially linked to cytokinesis was observed in 30 h ALH sip1 mutant clones. Despite that we cannot rule out the possible role of sip1 in apoptosis, these sip1 mutant clones had lower cell numbers and mitosis marker H3-P (Supplemental Fig. 10), which could help to explain the neurogenesis defect observed in adult brains with sip1 mutant neuroblast clones. However, further investigations will be needed to clarify whether SIP1 participates in cytokinesis during neuroblast proliferation or in apoptosis of neural cells and how such participation might affect neurogenesis. In contrast to Drosophila SIP1, the known characteristics of the mouse homologue of SIP1, Tuftelin-interacting protein (TFIP11), would imply that SIP1 might have a different and perplexing function. In another yeast two-hybrid screen, TFIP11 was identified as a binding partner for Tuftelin, one of the major proteins in enamel biomineralisation and possibly an RNA splicing factor 19 . Since there is no evidence that TFIP11 regulates neurogenesis-related processes to our knowledge, we cannot easily reconcile the functions of SIP1 and TFIP11 in the two systems. Interestingly, functional and developmental studies on the worm homologue of Drosophila SIP1, Septin and Tuftelin interacting protein 1 (STIP-1), may provide a solution for this conundrum. First, it is known that SIP1, TFIP11 and STIP-1 are functionally conserved proteins in worms, insects and mammals since embryonic lethality in C. elegans lacking stip1 can be rescued by overexpression of either Drosophila sip1 or human tfip11 13 . Second, STIP-1 is crucial for the early embryonic development and may be linked to cell division since worms with stip-1 knockdown exhibited arrested development and morphological abnormalities around the 16-cell stage 13 . Besides the STIP-1 studies in C. elegans, TFIP11 has also been linked to proliferation of cancer cells, as it was upregulated in non-small cell lung cancer (NSCLC); furthermore, knockdown of TFIP11 expression inhibited NSCLC cell proliferation, possibly due to cell cycle arrest and induction of apoptosis by key cell cycle-and apoptosis-related proteins 25 . Studies on TFIP11 in NSCLC together with the studies of STIP-1 in C. elegans may provide clues for deciphering the potential role of SIP1 in Drosophila central brain neurogenesis. Intriguingly, by studying RNA splicing co-factors in cell cycle and lineage progression in neuroblasts, sip1 RNAi knockdown caused underproliferation phenotype (sip1 was considered as a RNA splicing-related factor in Supplemental Fig. 11 of the Abramczuk study 26 ), which is similar to our MARCM experiments on sip1 mutants at the early larval stage (Supplemental Fig. 10). However, the subcellular distributions of TFIP11 and STIP-1 (in the nucleus of cultured cells) reported in previous studies 13,19 are very different from the localization of SIP1 (in plasma membrane/cytosol of neural cells) we saw in vivo (Fig. 6b-d and Supplemental Fig. 9). Therefore, future investigations into pre-RNA splicing, transcriptional regulation, apoptosis and Septin-directed cytokinesis will be crucial for elucidating the function and functional localization of SIP1 that control neurogenesis in the early developing larval brain.

Materials and methods
Generation of transgenic and sip1 mutant flies. Standard molecular biology techniques were used to generate UAS-transgene constructs, including CG7236, CG11030, CG11147, CG11149, DIP-θ, sip1, sip1::GFP, sip1ΔC::GFP and sip1ΔN::GFP. UAS-transgene constructs and two BAC genomic DNA clones (CH321-13P21 and CH321-86B19 obtained from Pacman Resources 16 ) were used to generate various transgenic flies by integrating DNA fragments into the VK33 docking site; performed by WellGenetics Inc. The sip1 GCFC mutant fly was generated by using the standard CRISPR-Ca9 method 17,18 to replace the coding sequence between 1240 and 1293 bp from the ATG site with a DNA fragment carrying the RFP-stop cascade, which should abolish the production of the GCFC domain; performed by WellGenetics Inc. Clonal analysis with MARCM and twin-spot MARCM. The generation, dissection, immunostaining and mounting of mosaic clones in adult brains have been described 12,33 . For MARCM experiments, mosaic clones were induced from NHL to 96 h ALH by heat-shock for 15-40 min. For twin-spot MARCM experiments, mosaic clones of embryonic-born adPNs were generated by heat-shock for 12 min. Primary antibodies used in this study included rat monoclonal antibody to mCD8 (1:100, Invitrogen), chicken antibody to GFP (1:800, Invitrogen), rabbit antibody to GFP (1:800, Invitrogen), rabbit antibody to RFP (1:800, Clontech), rabbit antibody to γ-aminobutyric acid (GABA; 1:100, Sigma), rabbit antibody to phospho-Histone H3 (H3-P; 1:200, Millipore), nc82 (1:100, Developmental studies hybridoma bank/DSHB) and mouse antibody to choline acetyltransferase (Chat; 1:100, DSHB). Secondary antibodies with different fluorophores, including Alexa 488, 546 and 647 (Invitrogen), were used at 1:800 dilutions in this study. Immunofluoroscence images were collected by Zeiss LSM www.nature.com/scientificreports/ 700 or 780 confocal microscopy and further processed using Adobe Photoshop. The plugin "Cell Counter" and the analyzing tool "ROI Manager" from Fiji ImageJ were used to count neuronal number and estimate the SIP1::sfGFP expression, respectively. One-way ANOVA with post-hoc Tukey test was used for statistical analysis in this study. www.nature.com/scientificreports/