Sox2 Transcriptionally Regulates Pqbp1, an Intellectual Disability-Microcephaly Causative Gene, in Neural Stem Progenitor Cells

PQBP1 is a nuclear-cytoplasmic shuttling protein that is engaged in RNA metabolism and transcription. In mouse embryonic brain, our previous in situ hybridization study revealed that PQBP1 mRNA was dominantly expressed in the periventricular zone region where neural stem progenitor cells (NSPCs) are located. Because the expression patterns in NSPCs are related to the symptoms of intellectual disability and microcephaly in PQBP1 gene-mutated patients, we investigated the transcriptional regulation of PQBP1 by NSPC-specific transcription factors. We selected 132 genome sequences that matched the consensus sequence for the binding of Sox2 and POU transcription factors upstream and downstream of the mouse PQBP1 gene. We then screened the binding affinity of these sequences to Sox2-Pax6 or Sox2-Brn2 with gel mobility shift assays and found 18 genome sequences that interacted with the NSPC-specific transcription factors. Some of these sequences had cis-regulatory activities in Luciferase assays and in utero electroporation into NSPCs. Furthermore we found decreased levels of expression of PQBP1 protein in NSPCs of heterozygous Sox2-knockout mice in vivo by immunohistochemistry and western blot analysis. Collectively, these results indicated that Sox2 regulated the transcription of PQBP1 in NSPCs.


Introduction
PQBP1 is a novel protein identified by yeast two-hybrid screening that binds to the polyglutamine (polyQ) tract sequence [1,2]. PolyQ sequence is shared by multiple transcription-related factors, such as TFIID/TATA-binding protein, androgen receptor, glucocorticoid receptor, octamer-binding factors, and CREBbinding factors, and by proteins related to a group of triplet-repeat diseases, which are called polyQ diseases, such asataxin-1, ataxin-2, atrophin-1/DRPLA protein, and huntingtin. These proteins belonging to the two groups were easily determined by putting the polyQ sequence into a BLAST search (http://blast.ncbi.nlm.nih. gov/Blast.cgi?PAGE = Proteins).The binding characteristics of PQBP1 suggests involvement of PQBP1 in the common pathology of multiple polyQ diseases [3,4].Although PQBP1 is a small protein of 265 amino acids, it possesses two protein-interaction motifs, the WW domain (WWD) and the C-terminal domain (CTD). PQBP1 interacts with RNA polymerase II (Pol II) or WBP11/NpwBP/SIPP through WWD [4][5][6] and with a U5 spliceosome factor U 5-15 kD through CTD [7,8].
PQBP1 is a nuclear-dominant nuclear-cytoplasmic shuttling protein that is engaged in RNA metabolism and transcription. PQBP1 functionally connects transcription and splicing when it is located in the nucleus [1,4,[7][8][9][10], but, in abnormal conditions, it shifts to the stress granule in the cytoplasm and probably regulates RNA metabolism [11]. Although the dynamics and functional roles seem to be essential for all types of cells, previous analyses have indicated an unequal distribution of PQBP1 mRNA and protein in different tissues and different cell types [1,12]. In embryonic mouse brain, our previous in situ hybridization study revealed that PQBP1mRNA is dominantly expressed in the periventricular zone region where neural stem progenitor cells (NSPCs)are located [12].
PQBP1 has been implicated in both neurodegenerative diseases and developmental disorders.PQBP1 has been shown to interact with the two polyQ disease proteins, ataxin-1 and huntingtin [3,4]. A European Consortium Study has revealed that mutations of thePQBP1 gene are linked to intellectual disabilities (ID) that are characterized by mental retardation, microcephaly, and short stature [13][14][15][16][17]. Because the expression pattern in NSPCs is critical for an understanding of ID and microcephaly inPQBP1 gene-mutated patients and of stem cell-related pathology in neurodegeneration, we investigated the transcriptional regulation of PQBP1 by NSPC-specific transcription factors in this study.
Especially, we focused on the role of Sox2 in the transcriptional regulation of PQBP1 in NSPCs. Sox2 is one of the most critical transcription factors for NSPCs. The stemness or pluripotency of NSPCs is mainly attributed to Sox2, which is clearly demonstrated in studies of induced pluripotent stem (iPS) cells [18]. The difference between totipotent embryonic stem (ES)/iPS cells and NSPC depends on the expression of Oct-3/4, which is expressed in ES/embryonic carcinoma (EC)/iPS cells [18][19][20][21][22], but it is rapidly downregulated during their differentiation by retinoic acidrepressive enhancer [19]. Actually, iPS cells were generated from NSPCs only with Oct-3/4 [23]. Sox2 forms a heterodimer with the POU transcription factors, including Oct-3/4 [24][25][26]. In this regard, the change in the Sox2 partner from Oct-3/4 to the NSPC-specific POU factor Brn2 or to another NSPC-specific transcription factor, such as Pax6, may switch on the expression of PQBP1 in NSPCs.
Sox2, which is a HMG-box-containing transcription factor that binds to the minor groove of DNA, forms a heterodimer with a POU transcription factor [27]. POU transcription factors are a group of AT-rich major groove-binding transcription factors, and they include Pit1, Oct-1, Oct-2, and Unc-86. In addition, the group includes Oct-3/4, which is a key transcription factor for ES cell differentiation [28] and iPS cell generation [18], and the other transcription factors.
Pax6 is expressed in apical and basal progenitor cells. Brn2 is highly expressed in NSPCs. Therefore, they could be the partner of Sox2 [27,29]. Although Oct-3/4 has been shown to be expressed at high levels in stem cells ranging from ES cells to epiblasts in the neural lineage [19][20][21][22], it is downregulated to an extremely low level in NSPCs [30]. Furthermore, a conditional knockout of Oct-3/4 induced by nestin-Cre did not show morphological changes in the brain [31], and this was distinct from PQBP1-conditional knockout mice (our unpublished results). Thus, Oct-3/4 could be excluded as the candidate. Oct-6 might be a third candidate that is expressed in the glial cell lineage from NSPCs [32,33] whereas the expression of PQBP1 is low in glial cells [1].
Hence, we employed the first 2 types of heterodimers (Pax6-Sox2, Brn2-Sox2), and screened for the genome sequence surrounding the PQBP1 gene that were homologous to the Sox2-Pax6 or Sox2-Brn2-binding consensus sequence for their binding activity and transcriptional activity. Consequently we identified 18 genome sequences that interacted with the NSPC-specific transcription factors. In utero electroporation into NSPCs revealed that some of them had cis-regulatory activities in vivo. Furthermore, we found decreased expression of PQBP1 protein in NSPCs of heterozygous Sox2-knockout mice. All these results in this study supported that Sox2 regulates the transcription of PQBP1 in NSPCs.

PQBP1 Protein is Highly Expressed in NSPCs
We have previously shown with in situ hybridization that mouse PQBP1 mRNA is expressed in mature neurons, but not in glia, in adult brains [1] and that PQBP1 is dominantly expressed in the periventricular zone in embryonic brain [12]. The levels of expression were clearly higher in embryonic brain than in adult brain [12]. At the beginning of this study, we reexamined the previous findings at the protein level by immunohistochemisty. In E15 embryonic brain, PQBP1was highly expressed in NSPCs, but it was also weakly expressed in differentiating neurons ( Figure 1). This finding reconfirmed the high levels of expression of PQBP1 in NSPCs that we found in our previous results.
Although the PQBP1 protein has been shown to shuttle between the nucleus and the cytoplasm [1] and to move to a specific cytoplasmic domain when cells form stress granules under stressful conditions [11], PQBP1 concentrates in the nucleus due to its nuclear localization signal sequence in normal conditions [1]. An immunohistochemical analysis with confocal microscopy again demonstrated the in vivo expression of PQBP1 in periventricular NSPCs of E15 mouse embryo ( Figure 2A).Moreover, PQBP1 was clearly colocalized with Sox2, which is a nuclear marker of NSPCs ( Figure 2A). Immunohistochemistry with an anti-PQBP1 antibody showed PQBP1 protein expression in embryonic brain (Figure 1), and this protein expression was downregulated during the differentiation of NSPCs in a western blot analysis that was performed with primary NSPCs before and after differentiation by culture on polyethyleneimine and poly-L-lysine ( Figure 2B).
The PQBP1 Gene is Surrounded by Multiple Sox2-Binding Sites As introduced, we employed the Pax6-Sox2 and Brn2-Sox2 heterodimers and screened for the genome sequence surrounding the PQBP1 gene. In the 57424000 to the 37524000 chromosomal region surrounding the PQBP1 gene, we found 132 sequences that were homologous to the Sox2-Pax6 or Sox2-Brn2-binding consensus sequence ( Figure 3 and Table 1). The criteria of consensus sequences are listed in Table 2. In addition, we found many binding consensus sequences for Sox2-Oct6, which will be discussed later.
From these sequences, we synthesized oligonucleotide probes and performed gel mobility shift assays. First, we checked the soundness of our method with positive controls of the cis-element ( Figure 4) and with GST-only proteins (Supplementary Figure 1). We next screened the 132 probes with GST-fusion proteins of the Brn2 or Pax6 DNA-binding domain and found68positive probes ( Figure 4 and Supplementary Figure 2B). The determination of positive probes was based on the radioactivity of the expected position of the band and on their statistical differences compared to the negative controls ( Figure 4). These were termed first-positive probes. Similarly, with the gel shift assay, we screened the candidate probes by the heterodimers of the full-length Sox2-Brn2 or Sox2-Pax6 proteins ( Figure 5 and Supplementary Figure 2A). The second screening selected 28 second-positive probes ( Figure 5). Through the comparison of first-positive and second-positive probes (Table 3), we selected double-positive probes, the binding of which to the Sox2 heterodimer was considered reliable. Consequently, we selected 18 probes (Table 4). Double-positive probes for Sox2-Pax6 were also positive for Sox2-Brn2, while double positive probes for Sox2-Brn2 were not always positive for Sox2-Pax6 (Table 3). Since both Brn2 and Pax6 prefer AT-rich sequences and Sox2 consensus sequences in minor groves are not extremely specific, such overlapped binding of the two heterodimers might have occurred.
In addition, we performed a gel shift assay with Sox2-Oct6 (Supplementary Figure 3). The results suggested that this complex could bind to part of the candidate cis-elements. However, Oct-6 expression in NSPCs has not been confirmed in previous studies, and our immunohistochemistry did not support Oct-6 expression in NSPCs (data not shown). In addition, as already mentioned we previously confirmed that PQBP1 is not highly expressed in glia compared to neurons [1], indicating that Sox2-Oct6 does not contribute to the transcriptional control of PQBP1. Thus, we did not use the complex in the following experiments.
In this study, we did not perform ChIP assay because the gel shift assays were multiplexed. However, ChIP assay might have strengthened our results, even though the bases of gel shift assay (supershift assay) and ChIP assay are overlapped.

PQBP1 is Regulated by Sox2
The double-positive sequences were distributed both upstream and downstream of the PQBP1 gene (Supplementary Figure 4). Four repeats of probe sequences in tandem were subcloned into the upstream sequence of the luciferase gene in reporter plasmids, and their cis-activities were tested in a luciferase assay. We cotransfected reporter and effector plasmids (Sox2+Pax6 or Sox2+Brn2) into P19 cells ( Figure 6A). P19 cells, which are equivalent to epiblasts, were selected as Sox2 and PQBP1expressing cells [1].
The luciferase assay showed obvious transactivation by the combination of transcription factors on the 18 cis-elements. The cotransfection of Sox2 and Pax6 showed remarkable transactivation with the #16A and #60 cis-elements ( Figure 6A). The cotransfection of Sox2 and Brn2 showed less remarkable transactivation, and most cis-elements showed definite higher  transactivation than that of negative controls ( Figure 6A). In these experiments, we used positive controls (PC) and negative controls (NC) possessing Sox2-Pax6 or Sox2-Brn2 consensus and mutated elements, respectively ( Figure 6A). Next, we tested whether the cis-element that was selected by the luciferase assay could induce transactivationin NSPCs in vivo ( Figure 6B). We performed electroporation of the reporter plasmids (pGreenFire1-PC-mCMV and pGreenFire1-16A-mCMV), which possessed enhanced green fluorescent protein (EGFP) gene cassettes downstream of the cis-element ( Figure 6A), into the brain of E14 mouse embryos and tested whether the reporter EGFP gene was expressed. As a marker of transfected cells, we cotransfected pLVSIN-CAG-pur-DsRed.When we used a positive control reporter plasmid (pGreenFire1-PC-mCMV), we observed EGFP signals in transfected cells (Ds-Red-positive cells) ( Figure 6B), indicating that the positive control cis-element for Sox2-Pax6 actually worked in vivo. Similarly, we observed green fluorescence in red cells when we transfected pGreenFire1-16A-mCMV ( Figure 6B and Supplementary Figure 5A). EGFP signal was detected in some cells outside of ventricular and subventricular zones, probably because NSPCs were differentiated before EGFP protein was completely degraded. Notably, the ratio between GFP and DsRed was clearly higher with pGreenFire1-16A-mCMV than with pGreenFire1-PC-mCMV, indicating a strong enhancer activity of 16A sequence. Meanwhile, in untero electroporation of a negative control plasmid, pGreenFire1-NC (Sox2+Pax6)-mCMV did not produce EGFP signals, indicating that the result was not artifact (Supplementary Figure 5B). All the results supported that the cis-element we selected possessed enhancer activity in NSPCs in vivo.

Sox2 Knockout Decreased PQBP1 in vivo
Finally, we tested whether a decrease of Sox2 affected PQBP1 expression in vivo. We used heterozygous Sox2-knockout (Sox2+/ 2) mice in which the Sox2 expression was almost 50% of control mice ( Figure 7A, B). The levels of expression of PQBP1 in Sox2+/ 2 NSPCs were about 50% of the control, judging from the western blot analysis of the primary NSPCs prepared from E18embryonic brains ( Figure 7A, B), and about 30% of the control, judging from the immunohistochemical staining of the embryonic brain (E18) (Figure 7C, D). These results strongly supported that Sox2 regulated PQBP1 expression in NSPCs not only in vitro, but also in vivo.

Discussion
In this study, we revealed the involvement of Sox2 in the transcriptional regulation of PQBP1in NSPCs. This conclusion was supported by multiple findings in vitro and in vivo. First, PQBP1 was coexpressed with Sox2 in NSPCs (Figure 1, 2). Second, Sox2 bound to multiple cis-elements that possessed transcriptional activities (Figure 3-6). Third, Sox2 haploinsufficiency reduced the levels of expression of PQBP1 in vivo (Figure 7). All of these findings indicated that PQBP1 was the downstream target of Sox2 in NSPCs.
Sox2 is a central player in various types of stem cells including ES cells. Therefore, one may ask why PQBP1 is not highly expressed in embryo before E11 [12] or in ES cells (our unpublished data). Actually Sox2 forms a heterodimer with Oct-3/4 in ES cells [34] and regulates toti-potency as shown in iPS cells [35]. However, the inner cellular mass that corresponds to ES cells does not need Sox2 for pluripotency, although it has been suggested that maternal Sox2 protein that remains in the cells compensates for the inactivation of Sox2 by gene targeting [36]. Considering that the expression of Oct-3/4 is extremely low in NSPCs [30,37], that the expression of Sox2 is found in epiblasts and NSPCs, and that Oct-3/4 is sufficient for reprogramming from NSPCs [23], the demand for Sox2 might be larger in NSPCs.
The Human ID diseases that are linked to PQBP1, such as Renpenning syndrome, Golabi-Ito-Hall syndrome, and Southerland-Haan syndrome, show multiple symptoms, such as microcephaly, short stature, lean body, and small testis [23]. The molecular pathways that result in each symptom are not yet understood. We have shown previously that the levels of expression of PQBP1 are correlated with the behavioral abnormalities. In PQBP1-knockdown mice in which the PQBP1 protein levels are reduced to 50% in the adult brain, we observed that anxiety-related cognition is impaired [38]. A consistent result was obtained in that Drosophila hypomorph mutants of PQBP1 homologs possessed a lower learning ability [39]. In addition, PQBP1 expression in NSPCs could be related to microcephaly, which is another phenotype of PQBP1-linked ID patients. With respect to these clinical aspects, the conclusion of this study that Sox2 controls the transcription of PQBP1 gene in NSPCs is critical.
There are various approaches to possibly modify the symptoms of PQBP1-linked ID patients. Because there are various alternative splicing isoforms of PQBP1 and the disease mutations of PQBP1 are complex, some isoforms that partially retain the functions of the full-length PQBP1 might be expressed in patients. Therefore, in analogy with Duchenne muscular dystrophy, increasing the levels of expression of partially active isoforms may alleviate the symptoms of patients. It would be also possible to enhance PQBP1 expression at transcription level in order to increase functional PQBP1. In this regard, the findings of this study might be useful for the development of a novel therapeutic approach that can be used to modulate the levels of expression of PQBP1 through Sox2.
Our findings were also important for developmental biology. Sox2 is now recognized as a key transcription factor that is involved in multiple neurological diseases [40]. This in turn indicates that Sox2 plays multiple physiological roles during development and in adulthood for normal brain morphology and function. Our results indicated thatPQBP1 is a target molecule of Sox2, which is part of Sox2 developmental functions.
The homozygous deletion of Sox2 is mostly lethal in mice, but the phenotype that is observed with other types of Sox2 mutations are thought to partially be mediated by PQBP1. An important finding from the Sox2 conditional depletion in the brain is neurodegeneration [41]. Those authors made 2 types of Sox2mutant mice. In one model involving b-geo knockin mice, they observed Sox2 expression in adult neurons and adult neurogenic regions and reduced cellular proliferation in the hippocampus and periventricular area [41]. They also observed in another mutated mouse model, in which a neural cell-specific enhancer was deleted, that Sox2 depletion in adult neurons resulted in the abnormal morphology of neurons and cytoplasmic inclusion body formation [41]. Considering these experimental results, PQBP1 could be a target of Sox2 not only during development, but also in adulthood. The role of PQBP1 in adult neurodegeneration could be our next target of investigation. Insufficiency of the Sox2-PQBP1 axis could be a cause of degeneration, and its upregulation might be used as a therapy of degeneration in the future as a regenerative approach, as suggested previously [42].  double positive probes selected for Luc assay : (Bold in Table 3)   4a  4b  4c  16a  16b  45   46  60  62  64  83  84   89  97  121  126  127  128 The double positive probes in Table 3 are listed. They were employed for luciferase assay (Fig. 6).   (Table 1) were subcloned into a reporter plasmid (pGreenFire1-mCMV), and their transcriptional activity was tested by cotransfection with effecter plasmids (Sox2+Pax6 or Sox2+Brn2) into P19 cells. Fold activation was calculated in comparison to the activity of the negative control reporter plasmid possessing a nonsense sequence as the ciselement. We simultaneously tested positive controls (PC) and negative controls (NC) that possess Sox2-Pax6 or Sox2-Brn2 consensus and mutated elements, respectively. Red and yellow bars represent transcriptional activity of PC and NC, respectively. (B) The transcriptional activity of the ciselement (P16A) in neural stem/progenitor cells in vivo was examined by in utero electroporation of the reporter plasmid. Electroporation was  performed on E14 and sampling was done on E15. The positive control reporter plasmid, pGreenFire1-PC (Sox2+Pax6)-mCMV, possesses a Sox2-Pax6 consensus sequence. pGreenFire1-16A-mCMV was generated by inserting the 16A oligpnucleotides:GTGAACCCTTTCAGATTTAGTGACGTAGCTTCA-CAAAGTGATTAA into pGreenFire1-mCMV. Confocal microscopy (LSM510META, CarlZeiss AG) with 40X water emersion lens was used to visualize the fluorescence. Green fluorescent protein signals were detected in NSPCs, even though the DsRed signals were weaker than the positive control, indicating that 16A possessed strong enhancer activity in vivo. doi:10.1371/journal.pone.0068627.g006 Figure 7. Sox2 regulates PQBP1 expression in vivo. (A) Western blot of primary-cultured neural stem progenitor cells from two Sox2+/2 mice and two littermate mice (WT) (E14). Sox2 and PQBP1 proteins were both reduced. (B) Ratios of western blot signal intensities (Sox2/internal control and PQBP1/internal control) were quantified. Asterisk indicates statistical differences (N = 6, p,0.05 by Student's t-test). (C) Immunohistochemistry of Sox2 and PQBP1 in Sox2+/2 and littermate (WT) mice (E18). Digital images were captured by an Olympus IX71 microscope. (D) Quantification of the intensity of Sox2 and PQBP1 immunostain signals from ventricular surface to cortical surface. Cortex at E18 were divided into 10 square areas from ventricular surface (0%) to cortical surface (100%) as shown in Figure 7C. The total signals were acquired in each square, background signals (mean signals in ventricles, N = 5) were subtracted from them, and the mean and SD values were calculated (N = 5). Asterisks indicate statistical differences (p,0.01 by Student's t-test). doi:10.1371/journal.pone.0068627.g007

Ethics Statement
This study was performed in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. The protocol was approved by the Committee on the Ethics of Animal Experiments of the Tokyo Medical and Dental University (Permit Number: 0130225A). All surgery was performed under ether of isoflurane anesthesia, and every effort was made to minimize suffering.