PiT2 regulates neuronal outgrowth through interaction with microtubule-associated protein 1B

PiT2 is a member of the inorganic phosphate transporter family, and is extensively expressed in the nervous system. It was found that loop7 domain of PiT2 is not required for retroviral recognition and transport function. The exact functions of loop7 remain poorly understood. Here we show that loop7 of PiT2 is necessary for the transport of PiT2 protein to the cell surface. Further, loop7 is also related to the outgrowth of neurite in Neuro2A cells interacts with the light chain 1 of microtubule-associated protein 1B (MAP1B). PiT2 with mutated MAP1B binding sites affect neurite outgrowth whereas Pi transport function deficient mutants of PiT2 do not. We also show that Drosophila dPiT interacts with microtubule-associated protein Futsch, and dPiT is crucial for the normal development of neuromuscular junctions (NMJs). These results indicate that PiT2 might participate in the regulation of neuronal outgrowth by interacting with MAP1B and independently of its Pi transport function in the nervous system.

To investigate possible functions of loop7 domain of PiT2 in the nervous system, we conducted immunofluorescence assays of Neuro2A cells transfected with PiT2 that lack loop7, and found that loop7 deletion affected the subcellular localization of PiT2 protein and neurite outgrowth in Neuro2A cells. To reveal the function of loop7 in PiT2, we performed yeast two-hybrid screening and identified microtubule-associated protein 1B (MAP1B) as a novel interactor of loop7 in PiT2. MAP1B is first synthesized and then cleaved to generate a heavy chain (HC) and a light chain (LC) 21 . As a cytoskeletal protein that regulates actin and microtubule dynamics, MAP1B plays important roles in axonal elongation and regeneration, neuronal migration, axonal guidance, dendritic spine morphology, as well as expression, trafficking and activity of neurotransmitter receptors 22,23 . Differentiation assay showed that MAP1B binding site mutants of PiT2 decreased the length of neurites in Neuro2A cells. In Drosophila, CG42575 (encoding dPiT protein) is homologous to human SLC20A2, and there is only one representative of MAP1 family: the futsch gene 24 . Futsch protein is cleaved similarly to MAP1 proteins in vertebrates 25 . Futsch is also implicated in neuronal development 26,27 . To dissect the neuronal function of loop7 domain in vivo, we generated transgenic lines that could be used to tissue-specifically overexpress dPiT with or without loop7. We performed co-immunoprecipitation and confirmed the interaction between Drosophila dPiT and Futsch. Immunochemical analyses showed that dPiT was crucial for the normal development of neuromuscular junctions (NMJs). This study reveals a novel function of PiT2 in neuronal outgrowth by interacting with MAP1B in vivo and in vitro.

Results
The loop7 domain is essential for PiT2 localization and might impact neurite outgrowth in Neuro2A cells. To get precise information about loop7 function in the nervous system, we first performed immunofluorescence assays of Neuro2A cells transfected with wild-type (PiT2-WT) or loop7 deletion mutant, in which residues 254-483 of PiT2 were deleted (PiT2-Δloop7). The PiT2-WT proteins were localized on plasma membranes in undifferentiated ( Supplementary Fig. S1a) and differentiated Neuro2A cells (Fig. 1a), but most of the PiT2-Δloop7 proteins were found in the cytoplasm, and aggregated in a specific region of the cytoplasm (Fig. 1b, Supplementary Fig. S1c). These findings indicated that loop7 might be necessary for trafficking of PiT2 protein to the cell surface. In differentiated Neuro2A cells transfected with PiT2-Δloop7, we observed that deletion of loop7 induced a decrease in neurite length compared with Neuro2A cells transfected with WT ( Fig. 1a,b,f). To further explore the biological function of loop7 in Neuro2A cell differentiation, we performed neuritogenesis assay. Following induction of differentiation by retinoic acid (RA) treatment, lengthening of Neuro2A cell neurites were detected. Knockdown of PiT2 by shRNA-PiT2 significantly decreased the length of the longest neurites by about one half compared with negative control (Fig. 1c-e,g and Supplementary Fig. S2). These results indicate that PiT2 might participate in the growth and development of the nervous system. Identification of MAP1B as a novel interaction partner of PiT2 by yeast two-hybrid screening. To search for interaction proteins involved in the subcellular localization or neurite outgrowth regulation of PiT2, yeast two-hybrid screening was performed. Residues 235-482 (loop7 domain) were used as bait (Fig. 2a), and was fused to the Gal4 DNA-binding domain. Through mating of the fetal brain cDNA library and Y187/ pGBKT7-loop7 we succeeded in screening approximately 400,000 independent clones. After selection of fetal brain cDNA library, 183 positive yeast clones showing His-reporter and Ade-reporter gene activity were selected. Further high-stringency selection and sequencing of the AD plasmid inserts led to the identification of two independent clones containing the light chain 1 (LC1) of MAP1B (Fig. 2b). The interaction between PiT2-loop7 and LC1 in yeast was reconfirmed by co-transformation of LC1 of MAP1B and loop7 of PiT2. The transformants show significant growth on SD/-Ade/-His/-Leu/-Trp selection agar plates, indicating an interaction between LC1 and loop7 (Fig. 2c).

PiT2 interacts with MAP1B in vitro and in vivo.
We substantiated the interaction between loop7 domain and LC1 by GST pulldown assay. The purified GST-PiT2-loop7 fusion protein, instead of GST alone, was able to pull down FLAG-LC1 fusion protein, indicating a direct association between loop7 and LC1 in vitro Yeast two-hybrid screen for the interacting protein of PiT2, and localization of MAP1B interaction site within loop7 of PiT2. (a) Schematic representation of PiT2, loop7 domain (residues 235-482, marked in red) was used as the bait for the yeast two-hybrid screen. (b) Schematic of the two yeast clones of MAP1B identified in the yeast two-hybrid screen. (c) Reconfirmation of the interaction between MAP1B and PiT2 in yeast. The transformants co-transformed with light chain of MAP1B and loop7 domain of PiT2 showed significant growth on SD/-Ade/-His/-Leu/-Trp selection agar plates compared with negative control. (d) Five C-terminal deletion mutants of loop7 were assayed for interaction with MAP1B using the yeast two-hybrid, only residues 235-402 interacted with LC1. (e) Loop7 truncation mutants were examined for interaction with MAP1B by the yeast two-hybrid assay, and only yeasts co-transformed with residues 356-402 and LC1 constructs showed growth on selection agar plates. (f) Alanine-scanning constructs designed to narrow the domains involved in interaction between residues 384 and 397. Mutations of residues 386-389 and 388-390 prevented the interaction between PiT2 and MAP1B. +, growth on stringent selection plates; −, no growth.
SCIENTIfIC RepoRts | (2017) 7:17850 | DOI:10.1038/s41598-017-17953-3 ( Fig. 3a and Supplementary Fig. S3a). Then full-length PiT2 and LC1 fusion protein expressing vectors were co-transfected into Hela cells. Lysates from co-transfected cells were immunoprecipitated with GFP antibody. Western blotting showed that GFP antibody was capable of pulling down LC1 and PiT2 fusion protein complexes in Hela cells (Fig. 3b,c and Supplementary Fig. S3b,c). We then carried out co-immunoprecipitation in mouse brain and Neuro2A cells lysates using LC1 antibody followed by Western blotting with PiT2 antibody, the results showed interaction between PiT2 and MAP1B (Fig. 3d,e and Supplementary Fig. S3d,e). After PiT2 knockdown, this interaction was weakened in Neuro2A cells ( Supplementary Fig. S4). In vivo, no interaction was detected in the supernatant brain lysates of PiT2 knockout mice ( Fig. 3d and Supplementary Fig. S3d).
MAP1B plays an important role in neurite extension during neuronal differentiation 22 . We performed co-immunoprecipitation in DMSO-or RA-treated Neuro2A cells. Compared with undifferentiated Neuro2A cells, PiT2 proteins co-precipitating with LC1 were roughly doubled in the differentiated Neuro2A cells (Fig. 4a,b and Supplementary Fig. S5a), suggesting that the interaction between PiT2 and MAP1B is affected by the differentiation of Neuro2A cells.
Mapping and verification of the MAP1B binding site on PiT2. To define the LC1 binding site in loop7 domain, we generated three constructs expressing truncated loop7 domain based on bioinformatics analysis ( Supplementary Fig. S6a) 28 . Yeast two-hybrid analysis revealed that deletion of either residues 356-482 or 235-402 prevented binding with LC1, while mutants with deletion of residues 235-355 and 403-482 retained interaction with LC1 (Fig. 2e, Supplementary Fig. S6b). Therefore, we constructed a series of mutants with C-terminal of loop7 truncated and found that only mutants containing residues 235-402 could interact with LC1 ( Fig. 2d, Supplementary Fig. S6c). To further localize the binding site of LC1 in PiT2, we generated six overlapping 4-residue alanine substitution mutations in the loop7 domain between residues 384 and 397. Among them, two alanine substitution mutants (residues 386-389 A and 388-390 A) failed to interact with LC1 ( Fig. 2f, Supplementary Fig. S6c), which indicate that residues 386-390 (YTCYT) are necessary for interaction between PiT2 and MAP1B.
To further verify the interaction sites, we performed a co-immunoprecipitation experiment in Neuro2A cells. HA epitope tagged PiT2-386-390A mutants were expressed in Neuro2A cells. Western blotting showed that LC1 antibody was able to pull down WT protein, but not 386-390A or Δloop7 mutant proteins ( Fig. 3f and Supplementary Fig. S3f).
Previous research results have demonstrated that LC1 is linked to the membrane localization of some ion channels and transmembrane receptor [29][30][31][32] . However, by immunofluorescence assay, we found that WT and 386-390A mutant PiT2 proteins localized to plasma membranes, and there were no considerable differences in the fluorescence intensity between WT and 386-390 A mutant ( Supplementary Fig. S1a,b).
Mutating the MAP1B binding site in PiT2 affects neurite length. Next, we conducted immunofluorescence assay and found that overexpressing PiT2 (386-390A) significantly decreased the length of neurites in Neuro2A cells (Fig. 4c,g). This indicates that mutating the MAP1B binding sites in loop7 leads to reduced neurite length.
We also evaluated whether transport function of PiT2 underlies the regulation of neurite outgrowth, and performed neuritogenesis assays in Neuro2A cells transfected with three PiT2 mutants found in IBGC families (S601W, R254*and V507Efs*2) 13,33,34 . The expression levels of 386-391 A, S601W, R254*and V507Efs*2 mutant in Neuro2 cells were similar (Supplementary Fig. S5b-c). Following induction of differentiation, Pi transport function deficient mutant PiT2-S601W (a missense mutation in C-PD domain) and PiT2-V507Efs*2 (deleted C-PD domain only) did not affect the neurite outgrowth in Neuro2A cells (Fig. 4d,f,g). The mutant PiT2-R254* (deleted loop7 and C-PD domain) as similar to PiT2-386-390A and PiT2-Δloop7, leading to significantly decreased length of neurites in Neuro2A cells (Fig. 4e,g). These results demonstrate that loop7 domain of PiT2 may participate in the regulation of neurite outgrowth in Neuro2A cells, while PiT2 lacking Pi transport function exerts no effect on neurite outgrowth.

The dPiT loop7 domain plays a crucial role in dPiT function in vivo.
Given that loop7 domain participates in PiT2 trafficking and regulates neurite outgrowth in Neuro2A cells, we examined if similar functions exist in Drosophila. Sequence comparison showed that the Drosophila genome contains the homolog of human SLC20A2, CG42575, which encodes dPiT protein. Drosophila dPiT is 38% identical and 56% similar to human PiT2 ( Supplementary Fig. S7b). We constructed dPiT loss of function mutants dPiT  and dPiT 15-1 by the CRISPR/Cas9 technology ( Supplementary Fig. S7). dPiT  and dPit 15-1 are frame-shift mutants carrying one base pair deletion at 62 th and 615 th nucleotide of the dPit gene, respectively. These mutants produced a truncated 43 and 191 amino acid peptides, respectively, and only 20 and 178 amino acids, respectively, in the C-terminal of this peptide are in common with WT dPiT protein. Heteroallelic or hemizygous mutants of dPiT which carry each of the mutation on one chromosome and the deficiency Df(3 L)ED4470 or Df(3 L)BSC817 that removes the entire dPiT gene on the other, were all embryonic lethal. Ubiquitous and neuronal overexpression of dPiT-GFP in dPiT loss of function mutant background by actin-Gal4 or elav-Gal4, respectively, rescued the lethality of loss of function mutant. However, ubiquitous or neuronal overexpression of dPiT-Δloop7-GFP in dPiT loss of function background by actin-Gal4 or elav-Gal4 could not rescue the embryonic lethality. These results suggest that dPiT is an essential gene for Drosophila development. The loop7 domain of dPiT is crucial for the function of dPiT.

Deletion of loop7 domain affects subcellular distribution of dPiT in Drosophila neurons.
Since aforementioned in vitro study showed that the loop7 domain played a crucial role in the trafficking of the PiT2, we then investigated the distribution of dPiT-WT and dPiT-Δloop7 in the neuronal system in vivo. Both dPiT-GFP and dPiT-Δloop7-GFP, when driven by elav-Gal4 in the wild-type background, were abundantly SCIENTIfIC RepoRts | (2017) 7:17850 | DOI:10.1038/s41598-017-17953-3 expressed in the cell body of Drosophila brain or ventral ganglions. While dPiT-GFP could also be detected in the axon and the terminal of NMJ, there were little distribution of dPiT-Δloop7-GFP in the axon, and it was hardly detectable in the NMJ (Fig. 5a,b' and 5e,f '). dPiT-GFP was found to be located in all parts of the sensory neuron system, including the cell body and dendrite branches. Nonetheless, dPiT-Δloop7-GFP mainly existed in the cell body and hardly observed in the dendritic branches of elav-Gal4/+ ;UAS-dPiT-Δloop7-GFP/+ flies (Fig. 5c,d'). These results indicate that loop7 domain affects subcellular localization of dPiT in neurons in vivo.
dPiT is required for normal the development of Drosophila NMJ synapses. To determine whether dPiT plays a role in synaptic development, we stained larval NMJs with anti-horseradish peroxidase (HRP) to label the neuronal membrane, and anti-cysteine string protein (CSP) to label synaptic vesicles. Knocking down dPiT in neurons or overexpressing dPiT-Δloop7-GFP was found to be associated with striking NMJ abnormalities as compared with controls or flies overexpressing of dPiT-GFP. The efficacy of the RNAi lines was validated by Western blot analysis with dPiT antibody that we generated and reduction of dPiT levels can be observed ( Supplementary Fig. S8e).
dPiT regulates NMJ development by interaction with Futsch. Microtubule-associated protein, Futsch is specifically expressed in Drosophila nervous system, and colocalizes with microtubule cytoskeleton in the well-studied Drosophila larval NMJ 24,26,35 . To test whether dPiT interacts with Futsch in the central nervous system, we performed immunoprecipitation using Drosophila brain. Western blotting of the immunoprecipitates exhibited an interaction between dPiT and Futsch in the brain (Fig. 6a,b and Supplementary Fig. S9).

Discussion
Previous studies and bioinformatics prediction showed that PiT2 is a highly hydrophobic protein consisting of 12 transmembrane domains (TMDs) and a large central intracellular loop (loop7) whose function remains unknown 14,20 . In this study, we found that MAP1B was a new interacting protein of loop7 domain. The interaction between PiT2 and MAP1B was demonstrated by yeast two-hybrid, GST pulldown and co-immunoprecipitation analysis. We found that the interaction was enhanced during the differentiation of Neuro2A cells. Overexpression of PiT2 with mutated MAP1B binding site resulted in a significant decrease in the neurite length of Neuro2A cells compared with wild type. Overexpression of Pi transport function deficient mutants PiT2-S601W and PiT2-V507Efs*2 did not affect neurite outgrowth in Neuro2A cells. These results suggest that PiT2 modulates neurite outgrowth independently of its Pi transport function. In vivo studies showed that dPiT possessed similar funtions in Drosophila. Drosophila dPiT interacts with Futsch, and dPiT is crucial for normal development of Drosophila NMJ synapses. Our data support the notion that loop7 domain of PiT2 is implicated in the growth and development of neurons by interacting with the adaptor protein MAP1B.
Most of the PiT2-Δloop7 proteins were localized to a specific region of cytoplasm ( Supplementary Fig. S1c). Previous studies have reported that MAP1B can mediate microtubular trafficking of Nav1.6 and 5-HT6R to the cell surface 29,30 . On the other hand, MAP1B interacts with CaV2.2 and 5-HT3A to reduce their expression in the plasma membrane and promoting their desensitization 31,32 . In this study, we found that mutations in residues 386-390 (YTCYT) impeded the interaction between PiT2 and MAP1B but did not affect its localization (Supplementary Fig. S1b). In vivo studies also revealed that dPiT-Δloop7-GFP fusion proteins predominantly existed in the cell body but not in axons, the branches of dendrites or the terminal of motor neurons in the elav-Gal4-driven UAS-dPiT-Δloop7-GFP flies (Fig. 5a-f '). Our results demonstrate that loop7 domain is required for membrane localization of PiT2 and interaction between PiT2 and MAP1B, but these two functions depend on different regions of loop7 domain.
MAP1B is highly expressed during early neuronal development 36,37 . MAP1B is principally expressed in neurons, oligodendrocytes, astrocytes, and their progenitor cells 38,39 . The expression pattern of MAP1B is similar to that of PiT2. MAP1B can regulate the dynamic balance between actin and microtubules, and it is essential for axonal growth, branching and nerve regeneration in developing nervous system [40][41][42] . In this study, we found that knockdown of PiT2 decreased the length of neurites in Neuro2A cells (Fig. 1d,e,g). The interaction SCIENTIfIC RepoRts | (2017) 7:17850 | DOI:10.1038/s41598-017-17953-3 between MAP1B and PiT2 was enhanced in differentiated Neuro2A cells (Fig. 4a,b) and abolishing the interaction decreased the length of neuritis (Fig. 4c,g). These findings suggested that the interaction between PiT2 and MAP1B might be involved in the differentiation of Neuro2A cells. Fly embryonic lethality of dPiT loss of function mutant illustrated that dPiT is an essential gene for fly development. We checked the NMJ phenotypes in dPiT loss of function mutant flies rescued with ubiquitous or neuronal expression of dPiT, respectively, and did not find any significant difference with wild type control in NMJ length and total bouton number (data not shown). This suggests that the crucial role dPiT plays development takes place in the neurons.
Immunoprecipitation assays showed that dPiT formed a complex with Futsch in Drosophila brain (Fig. 6a,b and Supplementary Fig. S9). Deletion of loop7 domain prevents dPiT from appropriate subcellular localization and affects normal protein function (Fig. 5). Moreover, dPiT and futsch mutants exhibit similar bouton growth phenotypes, and the phenotype of double mutants are similar to dPiT single mutants. Taken together, our results suggest that PiT2 regulates neuronal growth by interacting with microtubule-related protein MAP1B. Although Pi uptake of PiT2 in Xenopus laevis oocytes showed that loop7 domain is not required for Pi transport function 15 , there might be other regulatory mechanism for Pi uptake in the nervous system. PiT2 is a highly expressed inorganic phosphate transporter in the nervous system 10,43 . PiT2 could interact with actin and change their conformation to adapt to a changing ambient Pi concentration [44][45][46] . These results indicate that cytoskeletal proteins might play an important role in the regulation of PiT2 transport activity, and this might be related to the interaction between PiT2 and MAP1B in neuronal outgrowth regulation. In this study, we found that Pi transport function deficient mutant PiT2-S601W and PiT2-V507Efs*2 did not affect neurite outgrowth in Neuro2A cells (Fig. 4d,f,g). On the other hand, similar to PiT2-Δloop7, PiT2-R254* which also removes loop7 showed abnormal cytoplasmic localization and significantly decreased length of neurites in Neuro2A cells (Fig. 4e,g). These results show that PiT2 modulates neural outgrowth independently of its Pi transport function.
In summary, we identify a novel function of PiT2, which takes part in the growth and development of nerve cells. Furthermore, we find that PiT2 regulated the differentiation of nerve cells through interaction with MAP1B and independently of its Pi transport function. These findings might provide a novel mechanism that PiT2 regulates neural outgrowth, a process that might contribute to neuronal development.

Methods
Yeast Two-hybrid Assay. Yeast two-hybrid experiments were performed using the Matchmaker ™ Library Construction & Screening Kits (Clontech Laboratories, Inc., 630445). Briefly, the cDNA sequences encoding the human loop7 domain of PiT2 was amplified from KSM-hPiT2 vector 13 and subcloned into the pGBKT7 vector for use as "bait" in the yeast two-hybrid screen. A human fetal brain cDNA library as "prey" was purchased from Clontech (Clontech Laboratories, Inc., Mate & Plate ™ Library-Human Fetal Brain, 630469). The fetal brain cDNA library was screened by yeast mating, and then the mating mixture was spread onto complete medium lacking leucine, tryptophan, histidine and adenine (SD-Leu/Trp/His/Ade). In order to fully separate AD/library plasmid, candidate clones were restreaked on SD-Leu/Trp/His/Ade medium 2 times, and the β-galactosidase assay was performed using 5-bromo-4-chloro-3-indolyl-β-D-galactopyranoside (Clontech Laboratories, Inc., X-Gal, 8060-1). Plasmid DNA from each yeast colony was isolated and analyzed by polymerase chain reaction (PCR) and sequencing. The library inserts were identified using NCBI-blast search based on the DNA sequence. Bioinformatics analysis of the possible LC1 interaction sites within loop7 of PiT2 were performed using random forest algorithm 28 . Human MAP1B-LC1 cDNA (NM_005909.4) was amplified from pGADT7-MAP1B (2167-2468) vector (including residues 2167-2468 of MAP1B, which was identified in the screen) (Fig. 2b), and subcloned into pGADT7 vector. The pGBKT7-loop7 construct was used as the parental plasmid to generate the deletion and alanine substitution mutant constructs via PCR mediated mutagenesis 15,47 . The directed tests of the interaction between LC1 and loop7 mutants were performed using LiAc-mediated yeast transformation. The primers are listed in Supplementary Table S1.
Plasmids and Antibodies. Human MAP1B-LC1 cDNA was subcloned into p3×flag-CMV-7.1 and pEGFP-N1 vector. The full-length of wild type human SLC20A2 cDNA (NM_006749) was amplified from KSM-hPiT2 construct and subcloned into pEGFP-N1 vector. Full-length of human SLC20A2 cDNA with HA epitope tag sequence was subcloned into pCDNA3.1(−) vector, HA tag sequence was introduced into C terminus of PiT2 by PCR using two overlapped reverse primers. The pCDNA3.1-PiT2 construct was used as the parental plasmid to generate the mutant constructs via PCR-mediated deletion or site-directed mutagenesis. The pSIH-shRNA vectors containing either a sequence targeted to the mouse slc20a2 or a non-silencing control sequence (scramble) were used in RNA interference experiment. The primers are listed in Supplementary Table S2. For immunodetection, the following antibodies were used at the following dilutions: mouse anti-glutathione S-transferase antibody (ABclonal, AE001; 1:3000 for WB), mouse anti-GFP antibody (Proteintech, 66002-1-Ig; 1:5000 for WB, 1:100 for IP), mouse anti-flag antibody (MBL, M185-3L; 1:5000 for WB, 1:100 for IP),

Generation of UAS transgenic flies.
For overexpression studies, a UAS-dPiT-GFP construct was made by fusing the GFP with the C terminal of dPiT (NM_140184.4). Then we transformed w 1118 Drosophila with a UAS-dPiT-GFP fusion vector to generate transgenic flies. We also generated the UAS-dPiT-Δloop7-GFP vector. The insertion fragment was amplified from dPiT cDNA, connecting the sequence that encoding the 239 amino acids of N terminal to that of 183 amino acids of C terminal. Then we fused GFP with the C terminal of dPiT-Δloop7 fragment to generate the UAS-dPiT-Δloop7-GFP transgenic flies. The primers are listed in Supplementary Table S3. Mutagenesis of dPiT by CRISPR/Cas9 System. Drosophila Cas9/sgRNA system was used to mutagenize dPiT 49,50 . Two U6b-sgRNA plasmids (sgRNA1: gatgccaaaggcgagtacgaagg and sgRNA2: gatcgaaatccggccttgagcgg) were co-injected into nos-Cas9 transgenic blastoderm embryos to induce double-strand break at the first or the second exon of the dPiT (Supplementary Fig. S7a). We got two frame shift mutants, dPiT  and dPiT  , which was induced by sgRNA1 and the sgRNA2 respectively. dPiT  was the mutation with one bp deletion at the position of 62 th in dPiT cDNA. dPiT 15-1 also deleted one bp at the position of 535 th in dPiT cDNA ( Supplementary  Fig. S7b). The primers are listed in Supplementary Table S3.
Generation of endogenous dPiT::GFP by CRISPR/Cas9 System. We generated dPiT::GFP flies by CRISPR/Cas9 System 49,50 . We injected pCDF3-sgRNA-dPiT vector, which target a double-strand break (DSB) at the carboxyl terminal of dPiT, into syncytial blastoderm-stage embryos 49,50 . The sgRNA sequence was ACATGGGTGGGGCCTAAAGATGG. We also injected a circular double-stranded plasmid containing the GFP-coding sequence flanked by 1.7-and 1.4-kb homology arms from the dPiT locus into vasa-cas9 pCDF3-sgRNA-dPiT embryos. The donor template is designed to produce an in-frame insertion of GFP within the dPiT coding region, leading to a dPiT::GFP protein. The flies that expressed endogenous dPiT::GFP were screened by the GFP signal.
GST pulldown assay. GST tagged PiT2-loop7 fusion proteins were expressed by addition of isopropylthio-β-galactoside (IPTG, 0.1 mM, Sigma) in E.coli strain BL21 at 30 °C for 4 h. The HeLa cells were transfected with p3×flag-LC1 voter. Then glutathione-sepharose beads (Merck Millipore) were used to purify the GST fusion proteins according to the manufacturer's protocol and subsequently incubated with the HeLa cells lysates at 4 °C over night. The pulldown proteins bound to the beads were detected by Western blotting.

Immunoprecipitation. Expression vectors were transfected into Neuro2A cells using Lipofectamine 2000
(Thermo Fisher scientific). After culturing for 36 h, cells were lysed using IP lysis buffer (Beyotime) supplemented with cock-tail protease inhibitors (Roche). To detect endogenous interaction, 150 Drosophila (half male and half female adult flies emerging from the pupal cases within a week) heads or one newborn mouse brain was lysed using IP lysis buffer supplied with the Complete Protease Inhibitor Cocktail (Roche). Cell or tissue lysates were collected, and then centrifuged at 12,000 rpm, 4 °C for 10 min. Supernatants were immuno-precipitated with appropriate primary antibody over night at 4 °C. Then protein A-agarose beads (Merck Millipore) were added and incubated with the samples for another 2 h. For immunoprecipitation of Drosophila, lysates were incubated with appropriate primary antibody and Dynabeads ® Protein G (Thermo Fisher scientific). The beads were washed with IP lysis buffer three times. The immunoprecipitates were analyzed by Western blotting. Western Blotting. Lysates or immunoprecipitates were prepared using SDS sample buffer. Proteins were separated by 10% SDS-PAGE and transferred to nitrocellulose membranes. The membranes were blocked for 2 h at room temperature (RT) with 5% skim milk, and incubated with the appropriate primary antibodies overnight at 4 °C. Next day the membranes were washed three times with TBST buffer (20 mM Tris-HCl, 150 mM NaCl, 0.05% Tween-20, pH7.6), and incubated in HRP-conjugated secondary antibodies (1:20000, goat anti-rabbit or goat anti-mouse, Thermo Fisher scientific) for 2 h at RT. After rinsing three times, the proteins were detected by using SuperSignal ELISA Femto Maximum Sensitivity Substrate (Thermo Fisher scientific). The protein bands were quantitatively analyzed by employing the Image J software package (http://imagej.nih.gov/ij/). Immunofluorescence and Microscopic Analysis. For RNA interference experiment, Neuro2A cells were transfected with pSIH-H1-copGFP shRNA vectors, containing either a sequence targeted to PiT2 gene (shRNA-PiT2) or a non-silencing sequence (scramble). Transfected cells were fixed in 4% paraformaldehyde for 10 min and permeabilized with 0.5% Triton X-100 for 2 min at RT. After staining with DAPI (1:1,000, Sigma) for 5 min, cells were preserved at 4 °C. For exogenous expression experiment, transfected cells were fixed in 4% formaldehyde for 10 min at RT and permeabilized with 0.5% Triton X-100 for 2 min. Cells were incubated with the primary antibody overnight at 4 °C. The next day cells were incubated with the secondary antibody (Alexa Fluor 488-conjugated donkey anti-mouse IgG secondary antibody and Alexa Fluor 594 F(ab')2 fragment of goat anti-rabbit IgG(H+L) secondary antibody, 1:500, Thermo Fisher scientific) at 30 °C for 1 h. After staining with DAPI for 5 min, cells were preserved at 4 °C. For Immunochemical analysis of NMJ, third instar larvae were dissected in HL3 with all internal organs removed, followed by fixation in the 4% paraformaldehyde for 40 min. Dissection and antibody staining of third instar larvae were performed as previously described 51,52 . Immunofluorescent imaging SCIENTIfIC RepoRts | (2017) 7:17850 | DOI:10.1038/s41598-017-17953-3 was performed using an Olympus FluoView 1000 Laser Scanning Confocal Microscope mounted on an Olympus IX-81 inverted microscope. Image analysis was carried out using FV1000 Viewer and the Image J software. The length of the longest neurite in each Neuro2A cell stained green or NMJ length of Drosophila was measured from the swell of HRP staining to the terminal using Image J software package. All branches were calculated. For bouton size analyses, ImageJ 3.0 (NIH) was used to define anti-HRP-stained individual boutons. The software output reports the area for each bouton automatically. At least 16 NMJ4 terminals of different genotypes were analyzed. For quantification of GFP intensities in NMJ, staining signals were digitalized automatically using ImageJ and normalized to the average intensities in the corresponding HRP staining. 100 cells from each group in three independent experiments were captured for neurite outgrowth assay of Neuro2A cells.

Statistical analysis.
Experiments were repeated at least three times. Statistical analyses were performed with T-test for the comparison of two groups, and one-way ANOVA for the comparison of three or more groups. Quantitative data were presented as mean ± s.e.m., p values < 0.05 were considered to be statistically significant.