Identification of PBW or ACA domain module interactors by yeast two-hybrid screening

A partial region of mouse Nbea coding for the PBW domain module [GenBank: 158854037] (Asn2137-Tyr2936) was cloned into the bait yeast expression vector pB27 by Hybrigenics Services (France) (Fig 1B). pB27-PBW was transformed into the L40ΔGal4 yeast strain and Hybrigenics Services performed a Y2H screening of a mouse embryonic (E10.5 and E12.5) brain cDNA library (Hybrigenics) [28].

The Y2H screen for the ACA domain module of mouse Nbea was performed using the MATCHMAKER Two-Hybrid system (Clontech Laboratories Inc, CA, USA). The ACA domain module [GenBank: 158854037] (Met1-Met951) was cloned into the pGBKT7 plasmid (Fig 1B). This plasmid was co-transformed into the AH109 yeast strain together with a whole mouse embryonic (E12.5) cDNA pACT2 library [29], using the lithium acetate method [30]. After high stringency selection on synthetic dropout agar plates lacking leucine, tryptophane, and histidine (SD^---), positive clones were restreaked on new SD^---plates and grown for another 1 to 2 days. Only the prey plasmids of clones that survived the second high stringency selection round were extracted using the Prepease yeast plasmid isolation kit (Usb Corporations, OH, USA). The corresponding prey fragments were amplified by PCR and sequenced at their 5’ junctions. Sequences were then compared with the GenBank database using BLAST.

Construction of expression plasmids

A pcDNA3.1-FLAG-Nbea plasmid, containing full length mouse Nbea cDNA [GenBank: 158854037] and an N-terminal fused FLAG-tag, was used as a template for a PCR with the following primers: 5’-GCTCTCGAATTCCTTGCATTTATTGAAC-3’ and 5’-CTGAGCGGATCCTCAGTACCTGTTCTG-3’. This PCR fragment contained the cDNA coding for the NBEA DPBW domain module [GenBank: 158854037] (Leu1948- Tyr2936) flanked by an EcoRI and a BamHI restriction site and was cloned in frame into the pEGFP-C2 plasmid, resulting in the pEGFP-DPBW plasmid.

The sequence coding for the five arginines forming a classical nuclear localization signal (NLS) in the DPBW domain module were mutated into alanines using two site-directed mutagenesis PCRs on the pEGFP-DPBW template with the Quikchange Site-directed Mutagenesis Kit (Stratagene, CA, USA) (S1 Table). Mutations were sequence verified using Sanger sequencing.

Interaction network analysis

All interactors were uploaded into Cytoscape 3.2.0 to create a network of direct protein–protein interactions for the two separate domain modules [31].

The BiNGO plug-in was used to assess significant enrichment of gene ontology identifications (GO-ID) in the interactor list compared to the whole mouse genome (S2 and S4 Tables) [32]. The threshold for significance was set at 0.01. Non-redundant and brain tissue-related significant terms were selected to be of interest within the context of ASD.

On the 9^th of April 2015, a Pubmed literature search was performed to find publications linking any of the interacting proteins with ASD by combining the gene name with the term “autism” as a query. Full articles were then reviewed for a relevant link between the gene and ASD pathogenesis. All interacting proteins were used as a query in the Online Mendelian Inheritance in Man (OMIM) database.

The interacting proteins were identified as being PKA substrates if they were listed as such in the Phosphosite database [33].

Bio-informatics analysis

The mouse NBEA full length FASTA protein sequence [GenBank: 158854037] was submitted into several bio-informatics tools (NucPred, PSORT II, PredictNLS and NetNES 1.1) to assess its nuclear localization and look for consensus nuclear localization signals [34,35].

Cell culture and transfections

The mouse neuroblastoma cell line Neuro-2a (N2a) (CCL-131, obtained directly from American Type Culture Collection (ATCC), UK) or the human embryonic kidney 293 T cell-line (HEK293T) (CRL-3216, obtained directly form ATCC) were cultured in Dulbecco's modified Eagles medium supplemented with 10% fetal bovine serum at 37°C, 5% CO₂. For overexpression, N2a cells were transfected using TurboFect Transfection Reagent (Life technologies, CA, USA), while HEK293T cells were transfected using Fugene 6 transfection reagent (Promega, WI, USA), both according to the manufacturer’s protocol. For NBEA knockdown, N2a cells were transfected with the SMARTpool siGENOME Nbea (26422) siRNA (S1 Table, GE Healthcare, UK) using Lipofectamine RNAiMax (Life technologies) according to the manufacturer’s protocol. The Stealth RNAi siRNA Negative Control, Med GC (Life technologies) was transfected as a control. Cells were harvested 24h after transfection, except for experiments with NBEA knockdown, where cells were harvested after 72h [14]. NBEA knockdown or overexpression was verified by means of a quantitative RT-PCR on mRNA of transfected N2a cells.

When co-transfecting with other plasmids, knockdown was obtained by transfection with the previously described mU6Pro vector containing shRNA targeting the part of Nbea, coding for the C-terminal region, from amino acid 1863 until 2566 (mU6Pro-shRNANbea) [14]. The construct used for NBEA overexpression was pcDNA3.1-FLAG-Nbea, expressing full length mouse NBEA.

Luciferase reporter assay

The CBF1-dependent 4xwtCBF1 firefly luciferase reporter (4xwtCBF1Luc) has previously been used to asses NOTCH mediated transcription [36]. N2a cells were cotransfected with 390 ng of mU6Pro-shRNA-Nbea or pcDNA3.1FLAG-Nbea respectively, and with 320 ng of pCS2-NICD, 250 ng 4xwtCBF1Luc reporter and 40 ng pRL-TK plasmid. Control conditions were used for NBEA overexpression or knockdown, NICD overexpression and for 4xwtCBF1Luc overexpression by performing parallel transfections with the empty control plasmids or 4xmtCBF1Luc.

The Dual-Luciferase Reporter Assay System (Promega) was used to assess the effect of NBEA knockdown or overexpression on Notch-mediated transcription. To account for differences in transfection efficiencies, firefly luciferase activity was normalized the Renilla luciferase activity.

RNA extraction and quantitative reverse transcription PCR (qRT-PCR)

RNA of N2a cell cultures was extracted with the Nucleospin RNA Midi kit (Macherey-Nagel, Germany) according to the manufacturer’s protocol. Per experiment, an equal amount of RNA was reverse transcribed to cDNA using the iScript cDNA synthesis kit (Bio-Rad, CA, USA). Real-time qPCR was performed using iQ SYBR Green Supermix (Bio-Rad). 40 cycles of annealing/extension for 1min at 60°C were carried out with the MyiQ single color real-time PCR detection system (Bio-Rad). The relative amount of mRNA expression was calculated using the generalized qBase method [37]. In this method normalization is performed to the geometric mean of multiple reference genes. In this case two reference genes were used, peptidylprolyl isomerase A (Ppia) and growth arrest and DNA-damage-inducible alpha (Gadd45a). The primers used are given as S2 Table.

Nuclear extraction

The Nuclear Complex Co-IP kit (Active Motif, Belgium) was used according to the manufacturer’s protocol to obtain cytoplasmic and nuclear fractions of N2a cells.

Co-immunoprecipitation

Cells were harvested in lysis buffer (0.1% Triton-X100; 150 mM NaCl; 50mM Tris-HCl pH 7.4 and cOmplete, Mini, EDTA-free Protease Inhibitor cocktail (Roche, Germany)) and briefly sonicated on ice. After sonication cell debris was removed by centrifugation. Per sample, 20μl of a 50% bead slurry of protein G agarose beads (GE Healthcare) were used and washed twice with PBS before use. All incubations were performed at 4°C on a mechanical rotator. Protein G agarose beads were incubated with anti-GFP antibody (Roche) for 1h, after which they were incubated with PBS with 3% bovine serum albumine (BSA) (Sigma-Aldrich, MO, USA) for 1h. Beads were washed with lysis buffer after which cell lysate was added for 2h. Beads were washed five times with lysis buffer and proteins were harvested by resuspending the beads in sample buffer (50mM Tris-HCl pH 7; 10% glycerol; 2% SDS; bromophenol blue) compatible with immunoblot analysis.

Immunoblotting

A bicinchoninic acid protein assay (Life technologies) was used to determine the protein concentration of the samples. Samples were dissolved in SDS sample buffer. Depending on the molecular weight of the proteins of interest a 10% Bis-Tris gel (Bio-Rad) or 3–8% Tris-Acetate gel (Life technologies) was loaded with 25μg of each sample. Proteins were transferred to Protran Nitrocellulose membrane (Schleicher&Schuell, Germany) and incubated with antibodies against endogenous NBEA (1/3,000; rabbit polyclonal, homemade against the synthetic peptide TKVSDDILGNSDRPGS-Cys-KLHMBS (Eurogentec) as validated in [14] and in S1 Fig), GAPDH 14C10 (1/5,000; rabbit mAb, #2118S, Cell signaling technology), HDAC2 (1/2,000; rabbit polyclonal, sc-7899, Santa-Cruz Biotechnology Inc, CA USA), Myc (1/5000; mouse 9E10 mAb, home-made) and GFP (1/1,000; mouse mAb, 11814460001, Roche, Germany). Afterwards, membranes were incubated with HRP-conjugated secondary antibody (1/2,000; Dako, Denmark) and proteins were visualized with western blotting ECL detection reagent (Life technologies). Densitometry was performed using Image J analysis software (NIH) [38].

Confocal microscopy and image analysis

After washing with PBS, N2a cells were fixed for 30 min with 4% formaldehyde (Polysciences, PA, USA) in PBS at room temperature. Cells were washed three times with PBS, followed by permeabilization with 0.1% Triton-X100 (Acros) in PBS for 10 min. After blocking for 20 min with 5% goat serum (Dako) and 1% BSA (Sigma-Aldrich) in PBS (blocking buffer), cells were incubated with the primary homemade antibody for endogenous NBEA (1/1,000), diluted in blocking buffer, for 1h. After three washes, cells were incubated with a goat secondary Alexa594- or Alexa488-conjugated antibody (Life technologies) and DAPI (1/10,000) for 1h. Finally the cells were washed and mounted on microscope slides using Vectashield mounting medium (Vector laboratories, Canada). Cells were analyzed using the Olympus Fluoview FV1000 confocal laser scanning microscope and a 63x immersion oil objective.

The tool ‘Z project’ in ImageJ software was used to make a 2D Z-projection of a confocal Z-stack. All confocal images shown are single confocal planes, unless otherwise specified.

Quantification of nuclear EGFP fluorescence intensity was performed using Cell Profiler cell analysis software [39]. At least 300 transfected cells were measured per condition.

Statistical analysis

Data are represented as mean ± standard error of the mean (SEM). A student’s t-test was used, unless stated otherwise. All statistical tests were performed with 0.05 as the α-level of significance unless stated otherwise. With * P < 0.05, **P < 0.01 and ***P < 0.001.

The NBEA PBW domain module interacts with nuclear proteins

A Y2H screen for the two conserved domain modules ACA and PBW resulted in a list of 83 and 25 possible interactions, respectively (Fig 2, S3 and S4 Tables). Fourteen of these proteins (13%) have previously been reported within the context of ASD. In total, 31 genes (29%) are associated with neurological disorders. Five known PKA substrates (FLNA, ACTB, SOX9, MAP4K1 and GBF1) interacted with the domain modules of NBEA and three of these are linked with neurological disorders (FLNA, ACTB, SOX9). Both the PBW and the ACA domain interacted with Cortactin binding protein 2 N-terminal like (CTTNBP2NL). A lot of detected interactions were single hits, which highlights the need for further validation of these interactions in future experiments.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Y2H screening results for the ACA and PBW domain module.

The ACA and PBW domain modules of NBEA were used to screen for possible interaction partners. The results were visualized with Cytoscape software. Color coding was used to show genes related to ASD (red), to other neurologic disorders (orange) or no known association (blue). BiNGO enrichment analysis picked up an enrichment of ACA-interacting proteins for the cellular components: Cytoskeleton, Nucleus, Mitochondria, Collagen and Cytoplasm. The enrichment for cytoplasm is not shown in the figure. BiNGO analysis of PBW-interacting proteins only picked up the nucleus as an enriched cellular component. Some genes in this compartment also showed enrichment for the biological process of glial cell differentiation or Gliogenesis (purple framework). Proteins that are grouped as ‘Other’ were cytoplasmic or not picked up in the mentioned enrichments for cellular component. The rhomboid shape marks PKA substrates. The mutual interaction partner is connected with a green line.

https://doi.org/10.1371/journal.pone.0151954.g002

A search for GO-ID enrichment in the ACA or PBW interactor group compared to the whole mouse genome was performed with BiNGO. For the ACA interactor group a significant enrichment was found for the following cellular components: cytoplasm (53%; P = 2.3461E-7), nucleus (39%; P = 2.2586E-6), cytoskeleton (13%; P = 4.3418E-3), mitochondrial part (8%; P = 7.1712E-3), and collagen (4%; P = 1.5926E-4) (Fig 2, S5 Table).

For the PBW interactors there only seemed to be a significant enrichment for proteins present in the nucleus (92%; P = 2.8728E-15) (Fig 2, S6 Table). These interactors mostly have a molecular function in DNA binding (84%; P = 7.5905E-21) and/or transcription regulator activity (60%; P = 2.7411E-14). The PBW domain module also interacted with NUP153, which is a nuclear pore protein (Fig 2). Within the context of ASD it is of interest that there was a significant enrichment for biological processes in glial cell differentiation (17%; P = 9.8389E-7) and the NOTCH signaling pathway (13%; P = 5.7435E-5) (S6 Table).

Nuclear presence of NBEA

Neurobeachin is known to be expressed in the cytoplasm and localized near the trans-Golgi network in a punctate pattern. Since the NBEA PBW domain module and even the ACA domain module interacted with a large number of nuclear proteins, we first used bio-informatics to assess whether full length NBEA could be present in the nucleus. PSORT II and NucPred detected classical nuclear localization signals (NLS), although PredictNLS failed to detect any nuclear localization (Table 1, S1 Appendix). NucPred gave a nuclear localization score of 0.56, meaning there is a 70% chance of nuclear localization. No leucine-rich nuclear export signals (NES) were identified (S2 Appendix).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. PSORT II Computer prediction of putative NLSs in mouse NBEA.

https://doi.org/10.1371/journal.pone.0151954.t001

In N2a cells, low levels of NBEA (nuclear fraction of 19.1% ± 0.8) were detected in the nucleus, in addition to the previously described cytoplasmic localization using confocal microscopy (Fig 3A, S2 and S3 Figs, S1 Mov) [10]. This nuclear presence was confirmed by western blot of nuclear fractions (Fig 3B). Furthermore, the online database “Human protein atlas” supports these data, since immunostaining with a different NBEA antibody also shows NBEA expression in the nuclei of SH-SY5Y, human neuroblast, cells (S4 Fig) [40].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Nuclear localization of NBEA.

A) Confocal image of endogenous NBEA (yellow) in N2a cells. DAPI (cyan) was used to stain the nucleus. The separate images are shown in greyscale. B) Western blot showing endogenous NBEA at 327 kDa in cytoplasmic (C) and nuclear (N) fractions of N2a cells. GAPDH and HDAC2 were used as cytoplasmic and nuclear markers, respectively. * Aspecific band at 460 kDa.

https://doi.org/10.1371/journal.pone.0151954.g003

Stimulating the cAMP/PKA pathway in N2a cells with Forskolin and IBMX or inhibiting nuclear export with Leptomycin B did not increase the nuclear presence of NBEA (data not shown). The latter suggests that NBEA is not actively exported out of the nucleus by the exportin CRM1, consistent with the fact that no consensus NES was detected.

The DUF1088 domain in NBEA is important for nuclear localization

A penta-arginine sequence located in the DUF1088 domain was identified as a potential nuclear localization signal by NucPred and PSORT II (Fig 4A, Table 1, S1 Appendix). An EGFP-DPBW fusion protein was designed to determine its localization. The EGFP-DPBW fusion protein was detected in the cytoplasm and the nucleus as observed with fluorescence confocal imaging and western blot of nuclear and cytoplasmic fractions (Fig 4, S5 Fig, S2 Mov). A diffuse localization was observed, rather than a punctate pattern like full length NBEA. To assess if the predicted NLS in the DUF1088 domain is important for nuclear localization, the five present arginines (R) were mutated into alanines (A) (Fig 4A). Mutating this sequence resulted in a significant decrease of the confocal nuclear intensity (P = 0.047) (Fig 4B, S5 Fig, S3 Mov), demonstrating the importance of the NLS sequence for the nuclear localization of NBEA. A decrease of the nuclear intensity in nuclear and cytoplasmic fractions on western blot could be confirmed, but was not significant due to high variability of the overexpression levels between experiments (P = 0.063, Wilcoxon one-tailed signed rank test) (Fig 4C).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. The DUF1088 domain of NBEA contains a NLS.

A) EGFP-fused constructs were made containing the DPBW domain module. The NLS determined using bio-informatics is shown in an inset. For the EGFP-DPBW^MUT construct the arginines (R) of the EGFP-DPBW construct were mutated to alanines (A). B) Representative confocal image of N2a cells with overexpression of the EGFP-DPBW or the EGFP-DPBW^MUT fusion protein. DAPI was used to stain the nucleus. Quantification of nuclear intensity was based on at least 300 transfected cells per condition. Scale represents 5μm. C) Representative western blot of four experiments showing nuclear and cytoplasmic fractions of N2a cells transfected with pEGFP, pEGFP-DPBW or EGFP-DPBWMUT. For detection of the EGFP-fusion protein (170 kDa) an anti-GFP antibody was used (a-GFP). As a control for specificity for the fusion protein, no signal is seen for the unfused EGFP protein at this height, since it has a lower molecular weight of 27–30 kDa. HDAC2 and GAPDH were used as markers for nuclear and cytoplasmic fractions, respectively.

https://doi.org/10.1371/journal.pone.0151954.g004

Neurobeachin affects Notch-mediated transcription in N2a cells

Previously, Notch has been identified as a genetic interactor of the Nbea homologs rugose and SEL-2. In the Y2H screen the NOTCH1 intracellular domain (NICD) [Genbank: 224967065] (His2223-Leu2457) was identified to interact with the PBW domain module. This interaction was confirmed by co-precipitating overexpressed NICD-myc with EGFP-DPBW (Fig 5). Upon activation, NOTCH1 is cleaved and its NICD translocates to the nucleus to initiate transcription. Therefore, colocalization with NICD and the effect of NBEA expression on downstream NOTCH mediated transcription were investigated.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Overexpressed NICD interacts with EGFP-DPBW in HEK293T cells.

NICD-Myc co-precipitated with EGFP-DPBW in an immunoprecipitation (IP) using an anti-GFP antibody in HEK293T cells (n = 2). The input for the IP is shown on the left. An immunoblot (IB) for GFP shows EGFP-DPBW at the expected molecular weight of 170 kDa. The IB for Myc shows a signal for NICD-Myc at the expected molecular weight of 65 kDa. IP of EGFP did not lead to a co-precipitation of NICD-Myc, supporting a specific interaction between EGFP-DPBW fusion protein and NICD-Myc. kDa = kiloDalton.

https://doi.org/10.1371/journal.pone.0151954.g005

Overexpressed NICD will immediately translocate to the nucleus and only colocalized with the nuclear fraction of endogenous NBEA (Fig 6, S6 Fig, S4 Mov). However, this colocalization was not mutually exclusive, since the NICD nuclear localization was not punctate, but homogenous throughout the nucleus.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Overexpressed NICD partially co-localizes with NBEA in N2a cells.

Confocal image of endogenous NBEA (yellow) and overexpressed NICD (magenta) in N2a cells. DAPI (cyan) was used to stain the nucleus. The separate images are shown in greyscale.

https://doi.org/10.1371/journal.pone.0151954.g006

After nuclear translocation, the NICD forms a transcription activating complex with C-promotor binding factor 1 (CBF1) and Mastermind. Therefore, the previously described 4xwtCBF1 luciferase reporter was used to measure NOTCH activity. Overexpressing NICD-myc together with either overexpression (2.6-fold ± 0.5) or knockdown (0.7-fold ± 0.1; shRNA) of NBEA, resulted in a 50% decrease (P = 0.004) or increase (P = 0.003) in the Notch-mediated transcription of the luciferase reporter, respectively (Fig 7A and 7B).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. Notch-mediated transcription is affected by NBEA knockdown or overexpression.

A,B) NICD activity was measured with a luciferase reporter assay. The signal of the 4xwtCBF1 firefly luciferase reporter was measured and normalized to the renilla luciferase activity to control for transfection efficiency. The fold change of the normalized luciferase signal is shown compared to N2a cells with NICD stimulation (NICD) and without altered NBEA expression. N = 3/condition. A) N2a cells with or without NBEA KD (NBEA shRNA) were stimulated with overexpression of NICD. A representative western blot shows the 0.7-fold ± 0.1 KD that is obtained for NBEA (P = 7E-6). Equal loading of the western blot is shown by GAPDH. B) N2a cells with or without NBEA overexpression (FLAG-NBEA) were stimulated with overexpression of NICD. A representative western blot shows the 2.6-fold ± 0.5 overexpression that is obtained for NBEA (P = 0.001). Equal loading of the western blot is shown by GAPDH. C, D, E) qRT-PCR results are shown for Nbea and for four Notch target genes: Hes5, Hey1, Id3 and Id4. The fold change is shown relative to the condition with normal NBEA expression levels. C) The results for N2a cells with NICD stimulation together with or without knockdown of NBEA (NBEA shRNA) are shown. D) The results for N2a cells with NICD stimulation together with or without overexpression of NBEA (FLAG-NBEA) are shown. E) The results for N2a cells without NICD stimulation together with or without knockdown of NBEA (NBEA siRNA) are shown.

https://doi.org/10.1371/journal.pone.0151954.g007

To validate the results obtained with the luciferase reporter assay, we performed qRT-PCR to assess the expression levels of NOTCH transcriptional targets, namely Hes5, Hey1, Id3 and Id4. Combining NICD overexpression with knockdown of NBEA (0.3-fold ± 0.1; shRNA) indeed resulted in a significantly increased expression of all the addressed NOTCH targets (Hey1: P = 0.003; Hes5: P = 0.02; Id3: P = 0.02; Id4: P = 0.04) (Fig 7C), while NBEA overexpression (459-fold ± 161) resulted in a significantly decreased NOTCH-mediated transcription of Hes5 (P = 0.04) and Id3 (P = 0.04). For Hey1 and Id4 the differences were not significant (Fig 7D).

Since the Y2H screen for the PBW domain module also picked up HES5 and HEY1 as potential physical interactors, we assessed if knockdown of NBEA could also affect the transcription levels of NOTCH target genes independent of NICD overexpression. Knockdown of NBEA (0.15-fold ± 0.03; siRNA) resulted in decreased expression of Hes5 (P = 0.005) and Id4 (P = 7E-6), while no significant differences in expression could be observed for Hey1 and Id3 (Fig 7E). This means that the transcriptional repression of NOTCH targets by NBEA are mediated by NICD. Without NICD no (Hey1 and Id3) or even an opposite effect (Hes5 and Id4) is obtained.

Nuclear NBEA may regulate transcription

LRBA, the closest homolog of NBEA, has recently been shown to shuttle to the nucleus in mouse bone marrow cells after stimulation with LPS [44]. This homolog also contains the PBW domain module which is 76% identical to the module of NBEA. Just before the PBW domain module, a domain of unknown function (DUF1088) is conserved between both proteins. This domain includes a classical NLS consisting of a penta-arginine sequence, in NBEA as well as LRBA. By mutating this NLS we showed that the DUF1088 domain is important for the nuclear localization of the DPBW module. Despite several attempts, NLS mutagenesis of full length NBEA was unsuccessful due to its large GC-rich cDNA sequence. Therefore it is still plausible that other signaling might contribute to the nuclear localization of full length NBEA.

On average, 20% of total NBEA protein showed a punctate pattern in the nucleus of N2a cells, suggesting nuclear compartmentalization. NBEA might have a more pronounced nuclear localization after a specific cell stimulus, similar to LRBA. However stimulation of cAMP production, important for PKA signaling, did not increase nuclear localization of NBEA. Overexpression of NBEA in HEK293T cells, which do not express NBEA endogenously, did not result in any nuclear signal. Thus, a more pronounced nuclear expression will not only depend on cell stimulus, but also on cell type and developmental stage. Rugose, the shared homolog for LRBA and NBEA in D. melanogaster, has also been observed in some nuclei of neuroblasts delaminating from the ectoderm during stage 8–9 of embryogenesis [45]. Information obtained from the Human protein atlas further supports nuclear expression of NBEA in neuroblast cell lines [40].

Since NBEA is a large protein (327kDa), it will have to be actively imported into the nucleus. Active nuclear import can be mediated by three classes of importins and goes through nuclear pores. The Y2H screen for the PBW domain module also picked up the nuclear pore component 153 (NUP153) as a possible physical interactor, which is important for Importin-β mediated import [46]. Further studies will have to point out how NBEA is imported into the nucleus.

Recently, a second BEACH family member has been found to be implicated in ASD, namely WDFY3. WDFY3 also contains a PBW domain module and a CALL domain, but with very low sequence identity to NBEA. WDFY3 resides mainly in the nucleus, but does not contain a DUF1088 domain. Its function in the nucleus is unknown, but WDFY3 was still categorized into the transcription regulation network implicated in ASD pathogenesis [23,47,48]. Loss of WDFY3 in mice leads to increased proliferation of radial glia cells, increasing the brain cortical size [49].

Studies documenting networks important in ASD pathogenesis mostly pick up protein networks implicated in synaptogenesis and neural growth. But recently proteins involved in chromatin remodeling and transcription regulation have shown to be enriched protein networks for ASD as well [23,50]. Since, based on the Y2H screen, PBW interacted with a number of transcription regulators and NBEA is partially distributed in the nucleus, the effect of NBEA on transcription warrants further research.

NBEA is a negative regulator of Notch signaling

Here we provide support that NOTCH1 and NBEA can physically interact with each other and that NBEA expression levels can even affect NOTCH-mediated transcription. Studies in D. melanogaster and C. elegans have already suggested genetic interactions between the NOTCH pathway and the Nbea homologs, rugose and SEL-2, respectively [16,18,19]. Since in Drosophila Rg expression is not regulated by NOTCH-mediated transcription, interactions at the protein level with components of NOTCH signaling were suggested [16]. The effect of NBEA levels on NOTCH-mediated transcription is downstream of the NOTCH receptor activation, followed by its internal and external cleavage. Once NOTCH1 is cleaved, the NICD almost immediately translocates to the nucleus to initiate transcription. Therefore, there is a high probability that NBEA will interact with NICD mainly in the nucleus. Alternatively, NBEA overexpression may interfere with NICD nuclear translocation, for instance because they might use the same nuclear transport mechanism, namely Importin-β [51]. Interestingly, Yatim et al co-immunoprecipitated NBEA from nuclear extracts of a T-ALL cell line using NICD as bait [52]. These data confirm both the nuclear localization and the potential physical interaction of NBEA with NICD, as observed in our study.

Assessing the effect of NBEA expression on the downstream NOTCH-mediated transcription with the generalized CBF1-dependent luciferase assay, suggested for NBEA to be a negative regulator. There was an opposite effect between overexpression and knockdown of NBEA. This effect could be reproduced when studying specific NOTCH target genes with qRT-PCR, but with a somewhat variable outcome. The effect of NBEA overexpression on specific NOTCH target genes is less pronounced, compared to the effect on expression following NBEA knockdown. It can be expected that a more straightforward result is generated with the luciferase assay, since this is a general assay based on a luciferase plasmid, which is not endogenously expressed. In contrast, the NOTCH specific target genes are endogenously present in N2a cells and might be endogenously regulated at a more complex level, compared to the luciferase-construct. The complete composition of the NOTCH co-activator transcription complex is still not fully characterized and some negative regulators of NOTCH-mediated transcription have already been described. Furthermore, these target genes can be regulated at the level of epi-genetics, for example the methylation status of their promoter. A regional difference in the latter case could explain why not all NOTCH target genes behave in exactly the same way following altered NBEA levels.

Since NICD most likely physically interacts with NBEA, NBEA might interfere with the formation of the co-activating transcription complex of NICD. However, since NICD can also be regulated post-translationally and NBEA is an AKAP, NBEA could negatively regulate NOTCH-mediated transcription by affecting the phosphorylation status of NICD.

Although only canonical NOTCH target genes were assessed in this paper, it should be noted that non-canonical pathways could also be affected and contribute to a neuronal phenotype.

Until now, no ASD patients have been described with mutations in Notch genes, which may be because NOTCH is implicated in many biological processes during development and will cause more severe developmental disorders. Notch1 knockout mice already die during embryogenesis [53]. Mutations in human NOTCH1 can cause aortic valve disease and T-cell acute lymphoblastic leukemia [54]. However regulators of the NOTCH pathway that are highly expressed in the brain, like NBEA, might lead to less severe and tissue-specific symptoms that may be important for ASD etiology.

Recently, an integrated network analysis of de novo mutations in ASD patients has highlighted the NOTCH pathway as an important, enriched network. This network seemed to be common for ASD patients with and without intellectual disability (ID), while a network related to synaptic plasticity was only specific for ASD patients with ID [55]. In addition, three gene expression profiling studies of cerebellar or blood samples from ASD patients picked up an enrichment of the NOTCH pathway compared to controls [56–58]. In the central nervous system the NOTCH pathway is important for learning and memory, neuronal maturation and glial cell fate specification [59,60]. Conditional deletions of Notch1 in brain regions lead to precocious neuronal differentiation [59]. The NOTCH pathway is also known to be important in the cell fate decision between neurogenesis and gliogenesis. Gliogenesis also happened to be enriched for the Y2H PBW interaction list based on the following proteins: NOTCH1, ASCL1, SOX9 and SOX6. ASCL1, also known as MASH1, can regulate and is regulated by the NOTCH signaling pathway [61]. In contrast to NOTCH1, ASCL1 is a well-known proneural transcription factor and blocks gliogenesis [62]. HES1 and HES5, which are NOTCH1 transcription targets, can both repress ASCL1 expression and inhibit ASCL1 through protein-protein interactions. On the other hand ASCL1 can affect NOTCH signaling in neighboring cells by enhancing the expression of DELTA, a ligand of the NOTCH receptor [61]. SOX9 is a glial-specification transcription factor and has been shown to be a transcription target of NOTCH [63]. SOX6 expression can be regulated by SOX9 and has been shown to be important for cell specification to the glial lineage of oligodendrocytes, but is also essential for the maintenance of neural precursor cells by inhibiting neurogenesis [64,65].

A gene expression profiling study of lymphoblastoid cell-lines derived from ASD patients showed a significantly increased expression of SOX9 [56]. Post-mortem neuropathologic studies of persons with ASD often find a decreased number of GABA-ergic Purkinje neurons in the cerebellum and Bailey et al. also reported an increase of cerebellar Bergmann glia [66,67]. In general, a switch of cell-fate from neural to glial cells might contribute to ASD pathogenesis, in which the NOTCH signaling pathway can be important and where NBEA might be a relevant modulator.

The AKAP NBEA may bind PKA substrates

Based on the Y2H screens, the NBEA ACA domain module interacted with a number of validated PKA substrates (MAP4K1, ACTB, FLNA and GBF1), while the PBW module interacted with one (SOX9). It is known that AKAP-proteins are able to bind PKA substrates themselves, bringing the substrates in closer proximity of PKA. Since two other PKA substrates (CREB and CALPAIN2) have already been shown to have increased phosphorylation in vivo, due to NBEA haploinsufficiency, we assessed this for GBF1, ACTB and FLNA (data not shown)[26,68]. In N2a cells ACTB did not seem to be phosphorylated by PKA. GBF1 and FLNA showed equal expression after NBEA knockdown, but also equal PKA-mediated phosphorylation. Previously, we have demonstrated that haploinsufficiency of NBEA in mouse platelets affected a significant proportion of the PKA-related phosphoproteome [68]. Knockdown of NBEA in N2A cells did not produce such a pronounced effect (data not shown), suggesting that the chosen experimental settings may not be conductive to study the AKAP function of NBEA. However, it remains tempting to speculate that the ACA domain module plays a role in substrate presentation to PKA.

NBEA, neurotransmitter receptor trafficking and the cytoskeleton

The cytoskeleton was an enriched GO term, when analyzing possible interactors of the ACA domain module. This is in line with the suggestion that the Concanavalin A-like lectin domain is important for intracellular sorting. In addition this GO enrichment can be of interest for two reasons. Firstly, it has been suggested that NBEA might affect synaptic architecture through modulating the actin cytoskeleton [69]. Here we detected a direct interaction between the NBEA ACA domain module and β-ACTIN (ACTB). Secondly, BEACH proteins are suggested to be involved in (vesicle) trafficking, in which cytoskeletal proteins play an important part. NBEA already has been shown to affect postsynaptic trafficking of neurotransmitter receptors [12]. It has been described to bind to Glycine receptor β-subunit and Synapse associated protein 2 (SAP102), which is implicated in glutamate receptor trafficking. Knockout of NBEA resulted in reduced surface levels of glutamate and Gamma-aminobutyric acid A receptors (GABA_AR) [24,70]. Recent work found that the PKA-binding ability of NBEA can influence GABA_AR trafficking, but is not essential [71]. Here we discovered a possible interaction between the NBEA ACA domain module and the GABA_AR-associated protein (GABARAP). GABARAP regulates clustering of the GABA_AR at the plasma membrane and can bind directly to microtubuli and indirectly to the actin cytoskeleton [72,73]. Our findings suggest that NBEA might regulate the trafficking of the GABA_AR through binding with GABARAP and other cytoskeletal proteins, like ACTB and the microtubule network.

Previously a co-immunoprecipitation with full length NBEA in mouse adult tissue identified two tubulins, namely TUBB4A and TUBA4, as possible interactors [24]. Our Y2H analysis also picked up a tubulin, namely TUBA1A, further substantiating an important role for NBEA to be involved in the cytoskeleton.

The AKAP NBEA may bind other phosphatases and kinases

Based on the Y2H screens, CTTNBP2NL was the only common interactor of the ACA and the PBW domain module. Interestingly, CTTNBP2NL also interacts with two other previously described NBEA interactors, namely Striatin (STRN) and Zinedin (STRN4), forming the striatin-interacting phosphatase and kinase complex (STRIPAK) [24,74,75]. The STRIPAK complex has been described to be implicated in vesicle trafficking, dendritic spine formation and cell migration [75]. Although these specific interactions of NBEA with CTTNBP2BL, STRN and STRN4 have not been validated yet, the fact that they form a protein complex with each other and all have been picked up separately strengthens the possibility that these will be genuine interactors.

PP2A, present in the core of the STRIPAK complex, consists of three functional subunits, the structural subunit A, the catalytic subunit C and the regulatory or scaffolding subunit B. The subunit B can further be divided into three subgroups: B, B’ and B” [76]. The striatin family has been suggested to belong to a new B”‘ group, but this is still controversial [76,77]. In the Y2H screens the ACA domain did interact with other structural subunits of PP2A (PPP2R5E (B’ subunit) and PPP2R3C (B” subunit)), but we did not find any member of the striatin family to be interacting with a NBEA domain module. This raises the hypothesis that NBEA is able to interact with the serine/threonine (S/T) kinase PKA, but also with the S/T phosphatase PP2A.

A-kinase anchoring proteins (AKAPs) can form multi-protein complexes binding to other phosphatases and kinases as well [27]. In this way there can be a tight regulation of S/T phosphorylation and other posttranslational signaling in a specific cell compartment. In addition, other interacting kinases we picked up with the Y2H screens were BCKDK, PKM, MAP4K1, CDK4 and MAP2K2. NBEA might be an important scaffold for the formation of many different signaling complexes.