Ethics Statement

All procedures involving laboratory animals were performed according to the German laws of animal welfare and were approved by the ‘Landesamt fuer Natur, Umwelt und Verbraucherschutz NRW’ under the reference numbers 50.203.2-K 38, 19/05 and 8.87-51.05.20.10.106.

Generation of Neuronal-specific Tra2b Knock-out Mice and Mouse Handling

Neuronal-specific Tra2b knock-out mice were generated by cross-breeding Tra2b^fl/fl mice [21] with Tra2b^fl/+; Nestin-Cre^tg/0 mice. All mice were bread on a >96% C57BL/6N background. Nestin-Cre transgenic mice express Cre-recombinase under control of the rat Nestin promoter and enhancer and enable expression of Cre-recombinase in neuronal and glial precursors [48], [49], [50]. In a GFP reporter-line Nestin has been shown to be expressed in the ventricular and subventricular zone of the cortex [50]. For timed breedings the mating was allowed over night and females were checked for vaginal plugs in the morning. The day of plug detection was defined as 0.5 days postcoitum (dpc). For genotyping DNA was isolated from tail tip biopsies according to standard protocols and PCR using established conditions [21] (Dataset S4). For embryo or tissue preparation, euthanasia was performed by slowly increasing the carbon dioxide concentration in the animal housing.

(Immuno-)Histochemistry

Mouse embryos were isolated from pregnant females at indicated times by standard dissection protocols in PBS. Embryonic brains were fixated in 4% paraformaldehyde/PBS over night, dehydrated by immersion in ethanol series and embedded in paraffin. Histological stainings were performed according to standard protocols. In short, paraffin-embedded sections were deparaffinized in xylene, rehydrated by immersion in ethanol-series and shortly washed in PBS and water. For hematoxylin and eosin staining we used Mayer’s hematoxylin solution and Accustain eosin y solution (Sigma-Aldrich, Munich, Germany). Sections were incubated for 5 minutes in hematoxylin, washed for 15 minutes in water and stained for 30 seconds in eosin. For immunohistological stainings the rehydrated sections were subjected to antigen retrieval by boiling 3 times for 5 minutes in 0.01 M citrate buffer (pH 6). Sections were blocked for 45 minutes at room temperature in 20% serum and 1% BSA in TBS. Primary antibodies were applied over night at 4°C in TBS with 3% milk powder. For specific detection we used the following primary antibodies: rabbit-anti-Tra2b [23] 1∶8,000; rabbit-anti-Caspase-3 (R&D Systems, Minneapolis, USA) 1∶1,000; rabbit-anti-Ki67 (Abcam, Cambridge, UK) 1∶1,000. For detection we used a biotinylated anti-rabbit secondary antibody (1∶2,000 in TBS with 3% milk powder) for 30 minutes at 4°C and the vectastain ABC kit according to the manufacturer’s instructions (Vector Laboratories, Burlingame, USA). For colorimetric detection the DAB peroxidase substrate kit (Vector Laboratories, Burlingame, USA) was used. Images were recorded on a Zeiss Axio microscope and the Zeiss M2 imaging system (Zeiss, Jena, Germany).

Mouse Exome Microarrays

Mouse exon arrays were performed in the laboratories of Prof. A. Sachinidis at the Institute of Neurophysiology at the University of Cologne. Whole brain RNA samples of 16.5 dpc Tra2b^fl/+ and Tra2b^fl/fl; Nestin-Cre^tg/o mice were prepared using the Ambion WT Expression Kit (Ambion,Darmstadt, Germany) to amplify and generate sense strand cDNA with unbiased coverage of the transcriptome. cDNA was labeled and prepared using the GeneChip WT Terminal Labeling and Hybridization Kit (Affymetrix, Santa Clara, CA, USA). The array used was GeneChip Mouse Exon 1.0 ST Array (Affymetrix, Santa Clara, CA, USA). This array covers approximately 1 million exon clusters and 1.4 million probesets. Both gene level and exon level analysis were performed using R, Bioconductor package and Partek Genomic Suite tools. Preprocessing of exon array data included background correction using RMA [51], adjusting for GC-content, Quantile normalization and probeset summarization using mean and setting log probes using base 2. The probeset class, Gene extended was used. Preprocessing was performed by Partek Genomic Suite. Gene level analysis was performed in core extended mode. Transcripts showing very low expression signal values of less than 25 across arrays were eliminated and only 23,331 transcripts (annotated and non-annotated) were taken for statistical analysis. Moderated t statistics of linear model analysis [52] was employed to filter the differentially expressed transcripts. Threshold values of t score-p value of ≤0.05 and fold-change value 1.5 were used to generate a significant gene list. Partek Genomic Suite was used for statistical filtering of the alternatively spliced variants. Prior to alternative splicing calculation the probe sets with signals <3 (at log 2 scale) were excluded. The threshold values used for statistical filtering were set to p-value ≤0.05 and fold-change values of 1.5 for high confidence. Data obtained from array analyses are available at http://www.ebi.ac.uk/arrayexpress/with the accession number E-MTAB-2157.

Quantitative and Standard Polymerase Chain Reaction

DNA isolation from tail tip biopsies was carried out by standard protocols or using the Qiagen DNA Blood and Tissue Kit (Qiagen, Hilden, Germany). RNA was isolated from cells and tissues using the RNeasy Mini Kit and from fatty tissues using the RNeasy Lipid Tissue Mini Kit (Qiagen, Hilden, Germany). For all RNA isolations DNAse digestion was routinely performed. Concentrations were determined using a NanoDrop spectrophotometer (Peqlab, Erlangen, Germany) and the Quant-iT RiboGreen RNA reagent and kit (Life Technologies, Darmstadt, Germany). Reverse Transcription was performed using the QuantiTect Reverse Transcription Kit (Qiagen, Hilden, Germany). For routine PCR we used recombinant Taq DNA polymerase and for cloning procedures AccuPrime Pfx DNA polymerase (Life Technologies, Darmstadt, Germany). For quantitative real-time PCR we used LightCycler FastStart DNA Master SYBR Green I on a Roche Light Cycler 1.5 (Roche, Rotkreuz, Switzerland). Oligonucleotides and PCR conditions are given in dataset S4.

Minigene Splicing Assay

To assess splicing events we used the pSPL3-based exon trapping system from Life Technologies (Darmstadt, Germany). In short, the genomic region of the exon of interest with at least 500 bp of flanking intronic sequence was amplified from gDNA by PCR and cloned into the pSPL3 vector. Oligonucleotides are given in dataset S4. The vector contains a portion of the HIV-1 tat gene, an intron and splice donor and acceptor sites. The genomic region with the exon of interest is cloned into the intron, which is located in between the tat exons. The vector was subsequently co-transfected into HEK293T cells using the DharmaFect I transfection reagent (Thermo Scientific, Schwerte, Germany). Dependent on splice site detection the mRNA produced from the vector does either contain only the tat exons or additionally includes the exon of interest. The amount of each mRNA was determined by semi-quantitative RT-PCR. The splicing assay was performed while overexpressing GFP-TRA2B from pEGFP [23] or silencing Tra2b using specific siRNAs (Qiagen, Hilden, Germany).

Cell Culture

Murine embryonic fibroblasts were generated from 13.5 dpc embryos. Embryos were isolated from pregnant females by standard dissection protocols in PBS. Material was ground and washed through a 70 µm nylon filter with serum complemented DMEM. Cells were plated and incubated at 37°C, 5% CO₂ and 95% relative humidity. To block nonsense-mediated decay for detection of splicing isoforms cells were treated with 100 µg/ml emetine for 10 hours. Tra2b expression was depleted by stable transfection of a Cre-recombinase-GFP expressing vector into cells with the genotype Tra2b^fl/fl. Cre-recombinase was subcloned from pTriEx-1 kindly provided by Dr. Frank Edenhofer [53]. MEFs were transfected using the Mirus TransIT-LT1 transfection reagent (Madison, WI, USA) and grown under selective conditions.

Effects on tNasp expression were investigated in NSC34 cells [54]. NSC34 cells were a kind gift from Dr. Neil R. Cashman (Totonto) and have been produced by fusion of neuroblastoma cells with embryonic spinal cord cells. NSC34 cells display a neuron-like phenotype and are considered to resemble many aspects of motoneuron development [54]. HEK-293T cells [55] (human embryonic kidney) were obtained from ATCC. HEK-293T cells and NSC34 cells were transfected with plasmid DNA and/or siRNA using the DharmaFect1 Transfection Reagent (Thermo Scientific, Schwerte, Germany). Cells were harvested after 24 h, 48 h or 72 h for isolation of RNA and after 48 h and 72 h for the isolation of proteins. Cells harvested 72 h after transfection, were re-transfected after 48 h using initial conditions to maintain sufficient expression from a vector or sufficient knock-down from an siRNA, respectively.

Western Blot

Protein lysates from cells and tissue were prepared according to standard procedures. Samples were subjected to SDS-PAGE, transferred to PVDF blotting membranes and blocked for 2 hours at 4°C in TBST with 5% milk powder. For immunological detection of Tra2b we used rabbit-anti-Tra2b antibody 1∶2,000 in TBST with 2% milk powder over night [23]. Cdkn1a protein was detected by rabbit-anti-p21 (ProteinTech, Chicago, USA) 1∶1,000 in TBST with 2% milk powder over night. All quantification was normalized to beta-actin using mouse-anti-β-actin (Sigma-Aldrich, Munich, Germany) 1∶30,000 in TBST with 2% milk powder for 2 hours. For detection we used the horseradish-coupled secondary antibodies goat-anti-mouse-HRP (Jackson ImmunoResearch, Suffolk, UK) and goat-anti-rabbit-HRP (Thermo Scientific, Schwerte, Germany). As substrate for detection we used SuperSignal West Pico Chemiluminescent Substrate (Thermo Scientific, Schwerte, Germany).

Homozygous Loss of Tra2b in Neuronal and Glial Precursor Cells of the CNS Causes Perinatal Lethality in Mice

To generate neuronal-specific knock-out mice we crossbred homozygously floxed Tra2b mice (Tra2b^fl/fl) [21] with heterozygously floxed Tra2b mice carrying the Nestin-Cre transgene (Tra2b^fl/+; Nestin-Cre^tg/0). This Cre transgene is expressed under control of the rat Nestin promoter and enhancer [48]. Nestin expression is first activated at 10.5 dpc in the neural tube, the somites and in migrating cells of the neural crest. At 15.5 dpc it has been reported to be strongly expressed in the ventricular and subventricular zone of the cortex [49]. Using this breeding control mice (CTRL, Tra2b^fl/+ and Tra2b^fl/fl), heterozygous neuronal-specific knock-out mice (HET, Tra2b^fl/+; Nestin-Cre^tg/0) and homozygous neuronal-specific knock-out mice (KO, Tra2b^fl/fl; Nestin-Cre^tg/0) can be expected in a single litter (Fig. 1A). All four possible genotypes were equally distributed according to Mendelian law (Fig. 1B). Notably, all mice were born alive but KO mice died within the first hours of life. Only few KO mice survived up to a maximum of 36 hours, while HET mice had a normal life expectancy. Despite this drastic effect of Tra2b depletion, KO pups did not present major alterations in outer appearance (Fig. 1C). Previously we have shown that ubiquitous knock-out of Tra2b leads to early embryonic lethality in mice [21]. Here we demonstrate that the sole ablation of Tra2b in parts of the central nervous system causes perinatal lethality in mice and suggest that Tra2b is essential for neuronal development.

Download:

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

Figure 1. Conditional ablation of Tra2b causes perinatal lethality and disturbed cortical patterning in mice.

(A) Cross breading of Tra2b^fl/fl with Tra2b^fl/+; Nestin-Cre^tg/0 mice allowed generation of neuronal-specific knock-out (KO) animals as well as controls (CRTL) and heterozygous knock-out animals (HET) in one litter. (B) KO animals are born alive and all possible genotypes were detected according to Mendelian law (N = 113). (C) General development of conditional knock-out mice is not impaired as there are no gross morphological differences in embryo appearance. (D) Hematoxylin/Eosin staining of paraffin-embedded coronal sections at indicated developmental stages. KO animals but not controls or HET animals show ventriculomegaly of the third and lateral ventricles starting at around 14.5 dpc. Cortical layers are largely distinguishable at 14.5 dpc but cortical patterning and the ependymal lining of the lateral ventricle appears highly disturbed (black arrowheads) at 16.5 dpc in knock-out brains. (E) Immunostaining of Tra2b on paraffin-embedded coronal sections shows efficient downregulation of Tra2b protein in knock-out brains compared to controls and heterozygote animals. Cells of the ventricular and subventricular zones of the cortex show strongest decrease in staining intensity (black arrowhead). Scale bar equals 400 µm; ctx, cortex; th, thalamus; cpt, caudoputamen; cp, cortical plate; iz, intermediate zone; svz, subventricular zone; vz, ventricular zone; sn, septal nuclei; 3v, third ventricle; lv, lateral ventricle; pc, choroid plexus.

https://doi.org/10.1371/journal.pone.0089020.g001

Neuronal-specific Tra2b Knock-out Mice Show Severe Developmental Defects in the Brain and Aberrant Cortical Patterning

To understand the underlying cause of perinatal lethality we first histologically analyzed the brains of KO mice at various developmental stages using hematoxylin and eosin staining. Compared to control mice KO mice showed a dilation of both the third and lateral ventricles (ventriculomegaly) (Fig. 1D). The dilation of the ventricles had its onset at around 14.5 dpc and progressed until birth. Dilation at 14.5 dpc was still moderate and cortical patterning was comparable to control animals. At 16.5 dpc the lateral ventricles occupied a big part of the brain hemispheres and the cortical layers were no longer distinguishable. The ependyma which constitutes the ventricular lining appeared to be fully absent or destroyed in KO animals, even though the choroid plexus was present but being detached as a consequence of ventriculomegaly. Furthermore, late stage brain morphology (18.5 dpc) showed irregularities in the caudal regions of the septum or the anterior thalamic area. Heterozygous knock-out animals retained normal cortical patterning with normal ventricle sizes at all stages of development. This is in line with our finding of Tra2b expression being comparable to controls on brain sections (Fig. 1E, Fig. S1) and on Western Blots [17]. This suggests that a single gene copy of Tra2b can fully compensate protein function without producing an obvious phenotypic effect. The ventricular and subventricular zones of the developing mouse brain showed notably high Tra2b expression (Fig. 1E). However, Tra2b expression was strongly reduced in the ventricular and subventricular zones of KO animals. In general, brains of KO animals showed reduced levels of Tra2b expression but clearly retained localized centers of expression that likely comprise Nestin-negative cells (Fig. 1E, Fig. S1). This was in so far expected as the mosaicism of Cre expression is based on the natural expression pattern of the Nestin promoter. The absence or decay of the ventricular cortex layers in KO animals (Fig. 1D, Fig. S1) and the fact that these layers are loci of neuronal differentiation and neuronal outgrowth leading to cortex formation [56], [57] suggest that Tra2b plays a pivotal role in cortical development.

Brain Malformations are Initiated by Massive Apoptosis in the Cortex Followed by the Loss of Proliferative Potential in Neurogenic Areas

To further elucidate the course of the pathological progression we analyzed apoptotic events and the proliferative potential in KO versus control brains using immunostainings for cleaved caspase 3 (Casp3) and Ki-67. Localized presence of Casp3 indicating apoptotic events is a typical hallmark of neuronal development as each proliferation, differentiation and apoptosis orchestrate neuronal restructuring [58], [59]. KO brains showed drastically elevated levels of Casp3 indicative of apoptosis at 14.5 and 15.5 dpc but not at later stages of development (Fig. 2A). This primarily affected the cortical layers close to the lateral ventricles and the anterior thalamic region between the cavities of the third ventricle. Accordingly, the thalamic area presented morphological anomalies at later stages of development (16.5 dpc). The more dorsally located septal nuclei appeared largely unaffected, however (Fig. 2A, B, KO 16.5 dpc).

Download:

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

Figure 2. Brain malformations are initiated by massive apoptosis in the cortex.

(A) Immunostaining for Caspase-3 on paraffin-embedded coronal sections indicates prominent apoptosis in the proximal cortical layers and in the thalamic area of 14.5 dpc and 15.5 dpc KO embryos (black arrowheads). Remaining cortical tissue does not show apoptosis at 16.5 dpc and later stages (light arrowhead). (B) Immunostaining for Ki-67 shows initial decrease of proliferation at 14.5 dpc which is fully lost at 16.5 dpc in KO animals (black arrowheads). Control and HET animals retain strong Ki-67 signals in the proximal cortical layers at all indicated developmental stages. Scale bar equals 400 µm; ctx, cortex; th, thalamus; cpt, caudoputamen; sn, septal nuclei; 3v, third ventricle; lv, lateral ventricle.

https://doi.org/10.1371/journal.pone.0089020.g002

In concordance with these apoptotic events the expression of the proliferation marker Ki67 started to decrease at around 14.5 dpc in cells of the ventricular and subventricular zone and was fully lost at 16.5 dpc and all later stages (Fig. 2B). This temporal occurrence of events suggests proliferation loss being a direct consequence of initial apoptotic events ultimately leading to tissue loss and ventriculomegaly. At present we cannot discern whether cells affected by apoptosis are neuronal progenitor cells or differentiated neurons. However, cells are affected most in close proximity to the ventricle (Fig. 2A). In a GFP-reporter mouse line Nestin has been shown to be expressed in precisely these regions, the ventricular and subventricular zone of the cortex [50]. This might indicate that affected cells are mostly neuronal progenitors. The brain development of HET animals was not affected. In these brains we neither detected an increase of apoptosis, nor a loss of proliferative potential.

This demonstrates that haplosufficiency of Tra2b has no phenotypic consequence and rules out any secondary effect of the Nestin-Cre transgene on brain development. In contrast, depletion of Tra2b in neuronal precursors leads to ventriculomegaly and aberrant cortical patterning. This appears to be a direct result of massive apoptotic events in ventricular and subventricular cortical layers and subsequent loss of the proliferative potential in these neurogenic areas. These findings clearly point out a pivotal role of Tra2b in neuronal development and survival.

Tra2b Knock-down Efficiency is Variable between Individuals

Though all neuronal-specific KO mice presented ventriculomegaly, disturbed cortical patterning and perinatal lethality, the severity of the brain phenotype was variable between animals. Using whole brain RNA we first analyzed expression levels of Tra2b at different developmental stages. By real-time RT-PCR we found that Tra2b was effectively down-regulated in brains of 16.5 dpc, 18.5 dpc and PND 1 KO animals (Fig. 3A). As whole brain RNA has been used, residual Tra2b expression is likely a result of Nestin-negative cells that do not express Cre recombinase. As expected, HET mice showed a moderate but significant down-regulation of Tra2b transcripts. In control animals the total expression of Tra2b in the brain was largely constant between 16.5 dpc and birth (Fig. 3A).

Download:

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

Figure 3. Bona fide Tra2b-related splicing events can be detected in vivo in conditional Tra2b knock-out mice.

(A–G) Quantitative real-time PCR using whole brain RNA of control, HET and KO embryos at 16.5 dpc. All mRNA levels have been normalized to Gapdh. (A) Total Tra2b expression levels are effectively reduced in KO animals at 16.5 dpc, 18.5 dpc and PND1. (B) Expression levels of both the functional and non-functional Tra2b isoforms are reduced at 16.5 dpc. (C) Individual expression levels of total Tra2b largely behave accordingly to the respective genotype but are subject of variation within each group. Horizontal lines indicate the mean of each genotype group. The dashed line indicates a threshold value of 0.4. (D) Expression levels of the 3R and 4R isoforms of the Mapt transcript. Transcript levels of the 4R isoform are reduced in KO brains. (E) Expression levels of the LCbII and LCbIII isoforms of the Cltb transcript. Transcript levels of the LCbII isoform are reduced in KO brains. (F) Expression levels of the somatic and the testis-related isoform of the Nasp transcript. Transcript levels of the testis-specific T-isoform are drastically reduced by −5.5-fold in KO brains. (G) Expression levels of the functional and non-functional isoform of the Tra2a transcript. Transcript levels of the non-functional isoform are drastically reduced by −4.0-fold and levels of the functional isoform are increased by +1.4-fold. Mapt, microtubule associated protein tau; Cltb, Clathrin light chain b; Nasp, Nuclear autoantigenic sperm protein; Tra2a, Transformer 2 alpha; a.u., arbitrary units; error bars indicate the s.e.m.; significance levels are *p<0.05, **p<0.01, ***p<0.001 (Student’s t-test); N is given by numbers within bars.

https://doi.org/10.1371/journal.pone.0089020.g003

Tra2b is known to autoregulate its expression level via a splicing feedback-loop by including cassette exon 2 into the transcript [60]. Exon 2 contains a premature termination codon rendering the transcript non-functional and thus decreasing the Tra2b protein concentration. Using isoform-specific real-time PCRs we found a significant down-regulation of both the functional (−78%) and non-functional (−88%) Tra2b isoforms (Fig. 3B). In general the down-regulation of the non-functional isoform was stronger than for the functional isoform. Interestingly, in HET mice the levels of functional transcripts were only mildly reduced (−24%) while non-functional transcripts were very strongly reduced (−61%). These data suggest that the auto-regulatory splicing feedback-loop is functional in mice [17]. The only mildly altered levels of functional Tra2b transcripts are consistent with survival and normal brain development in HET mice. This suggests that loss of a Tra2b gene copy can be compensated to a great extent via this auto-regulatory mechanism.

Though average Tra2b knock-out was effective in KO mice we found individual differences between single animals while comparing animals of all genotypes at 16.5 dpc (Fig. 3C). Despite transcript levels were on average gradually reduced according to the genotype (horizontal lines), we found single KO animals with Tra2b transcripts hardly lower than in control animals. Variability and mosaicism in Cre expression throughout the brain of different individuals might explain that effect. To still allow functional analysis of splicing events in Tra2b knock-out mice we chose KO mice with relative Tra2b/Gapdh transcript levels of lower than 0.4 (Fig. 3C, dashed line).

Physiological Splicing Events of Bona Fide Tra2b Targets can be Reconstituted in Neuronal-specific Tra2b Knock-out Mice

To check whether relative Tra2b/Gapdh levels lower than 0.4 have an impact on physiological splicing events in the brain we quantified the isoform splicing of bona fide Tra2b target transcripts. Using isoform-specific real-time PCRs we analyzed the splicing of the well-known splicing targets microtubule associated protein tau (Mapt) exon 10 [25], [26], [27], the vesicle coating protein clathrin light chain b (Cltb) exon 5 [15], [29], the nuclear autoantigenic sperm protein (Nasp) exon 7 and the Tra2b paralog transformer 2 alpha (Tra2a) exon 2 [17], [61]. For Mapt (Fig. 3D), Nasp (Fig. 3E) and Tra2a (Fig. 3F) inclusion of the respective cassette exon into the transcript is promoted by Tra2b. Accordingly, we found a significant decrease in the Mapt 4R, Nasp-T and non-functional Tra2a isoforms. This demonstrates that these Tra2b-dependent exons are more frequently skipped in KO mice. For the Cltb transcript Tra2b promotes skipping of exon 5 [15], [29]. Accordingly, we found a significant increase in the exon 5 containing LCbII isoform in KO mice (Fig. 3G). In summary, these data demonstrate that our neuronal-specific Tra2b knock-out system can reconstitute well-known splicing events in the developing murine brain. Consequentially, we sought to perform an exome-wide screening for alternative splicing events in the developing central nervous system using whole exome array technology.

Identification of Tra2b Targets via Mouse Exon Array Analysis

To identify novel splicing targets for Tra2b we performed exome analyses of four CTRL and four KO whole brain RNAs from 16.5 dpc pups using Affymetrix mouse exon 1.0 ST arrays. By quantifying total Tra2b levels we assured a strong Tra2b knock-down with normalized expression levels of lower than 0.4 in the analyzed samples (see Fig. 3C). From the mean values of the CTRL and KO group we calculated global changes in the expression of whole transcripts and exon-specific differences, which would point out differences in splicing inclusion (PSI, percentage of splicing inclusion). Most of the identified exons showed no or very little changes in splicing inclusion (Fig. 4A, gray bars). As a threshold for PSI changes we filtered out exons with less than ±1.5-fold up- or downregulation, which excludes about 97% of the identified exons. Furthermore, exons with significance levels of worse than 0.05 were filtered out, still yielding a list of 1,006 differentially included exons (Fig. 4B, Dataset S1). Unexpectedly, we found multiple exons from one and the same gene being identified as alternatively spliced (132 exons in total). We compared the list of 86 differentially expressed transcripts (Dataset S2) with those genes called with multiple exons and found that the 132 alternatively spliced exons were matching with 54 out of 86 differentially expressed genes. We analyzed a subset of double-identified transcripts including the transcription factors Eomes, Tcf4 and others and found general changes in expression but no exon-specific changes (Fig. S2). This indicates that these exons being identified to be alternatively spliced are truly artifacts caused by misinterpretation of altered transcript levels. Thus, exons belonging to this group were excluded from our candidate list (Fig. 4B). The generally very high number of alternative events and the strong overlap with differentially expressed transcripts is probably indicative of poor data quality and a high error rate which strongly increases the number of false positive exons. We propose that the strongly altered brain morphology and misdevelopment in the KO mice and the according changes in cell type and transcriptome composition cause a large number of artifact results. Furthermore, mosaicism of Cre recombinase expression leads to two distinct cell populations in the KO brain. First, surviving cells in which Tra2b was depleted and show altered splicing patterns and, second, cells that express Tra2b and show normal splicing. In such a mosaic brain the latter cell population masks the splicing effect present in KO cells. That phenomenon appears to hamper our efforts to reliably call alternative splicing events.

Download:

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

Figure 4. Mouse whole exon array analysis reveals Tubd1 exon4 and Sgol2 exon4 as in vivo targets of Tra2b.

Whole brain RNA of 4 CTRL animals and 4 KO animals was analyzed on mouse exon array. (A) The inclusion ratio (PSI, percent splicing inclusion) of each identified exon is defined as [PSI_KO]/[PSI_CTRL]. PSI distribution reached from ∼0.2 until ∼4.0. (B) Initial filtering strategies comprised exclusion of PSIs between 0.66 and 1.5 (grey bars) as well as restriction to p-values smaller than 0.05 which yielded a total of 1,006 exons. Exons associated with transcripts identified as being transcriptionally up- or downregulated were excluded from analysis. Ranking of those was further refined using large PSI values and considering presence of putative Tra2b binding sites (AGAA-motifs). Thereby, exons had to contain at least a single AGAA-site and a AGAA-frequency higher than 1.5. (C,D,G,H) Semi-quantitative RT-PCT on whole brain RNA was carried out using isoform specific primers for Sgol2 FL (C), Sgol2 Δ4 (D), Tubd1 FL (G) and Tubd1 Δ4 (H) confirming splicing events identified on the microarray. All isoform expression levels were densitometrically measured and normalized against Hprt (E,F,I). The Tubd1 Δ4 isoform could not be detected using whole brain RNA, as skipping of exon4 introduces numerous premature termination codons leading to nonsense-mediated decay of the transcript (H). Treatment of wt and Tra2b-depleted murine embryonic fibroblasts with emetine successfully inhibited NMD and the Tubd1 Δ4 isoform was detectable in Tra2b-depleted cells only (J). FL, full length; Δ4, transcript lacking exon 4, (−) PCR negative control; a. u., arbitrary units; error bars show the s.e.m.; significance levels are *p<0.05, **p<0.01, ***p<0.001 (Student’s t-test).

https://doi.org/10.1371/journal.pone.0089020.g004

As an attempt to counteract this problem we excluded exons that do not contain any AGAA binding motif (Fig. 4B). Considering that such motif occurs statistically once in a 256 bp stretch we calculated the binding site frequency (F) as the number of binding sites multiplied with the exon length in bp divided by 256. Exons with F <1.5 were excluded resulting in 481 remaining exons (Fig. 4B, Dataset S3). To verify candidate exons we performed isoform-specific semi-quantitative RT-PCRs for a subset of 20 selected candidates (Fig. S2, and data not shown). We were largely unable to validate splicing events of candidate exons as the expression of both respective isoforms was either indifferent between CTRL and KO animals (e.g. Slu7, Nrg1) or one of the two isoforms was not detectable in the brain at all (e.g. Sox6) (Fig. S2C). However, splicing events of the Shugoshin-like2 and Tubulin delta chain 1 transcripts could be validated in our candidate screening (data not shown). Therefore, both transcripts were selected for further analysis via RT-PCR and minigene splicing assay. Results obtained from microarray analyses imply that the majority of data output from the mouse exome array are false positive artifacts based on either general transcript loss due to cell loss and misdevelopment or drastic changes in tissue and cell-type distribution in the severely affected KO brains. We propose employing a more homogeneous knock-out cell system is necessary for a successful exome-wide screening approach – especially to reliably detect changes in exon splicing.

Shugoshin-like 2 (Sgol2) Exon 4 and Tubulin Delta Chain 1 (Tubd1) Exon 4 are in vivo Targets of Tra2b

Using isoform-specific semi-quantitative RT-PCR we were able to reproduce the splicing pattern of Shugoshin-like 2 (Sgol2) and Tubulin delta chain 1 (Tubd1). While the isoform of Sgol2 including exon 4 only showed a modest and non-significant decrease between CTRL and KO brains, the Δ4 isoform significantly increased by 2.75-fold thereby resembling the 2.7-fold loss of exon inclusion detected on the microarray (Fig. 4 C–F, Dataset S3). These data demonstrate that the inclusion of exon 4 into the Sgol2 transcript is promoted by Tra2b. Similarly, we detected a splicing response for exon 4 of the Tubd1 transcript. The isoform including exon 4 showed a significant decrease by 1.46-fold in the brains of KO animals compared to CTRL (Fig. 4G–J) resembling the 1.7-fold loss of exon inclusion detected on the microarray (Dataset S3). Surprisingly, we were unable to detect the Δ4 isoform of Tubd1 in brains of either CTRL or KO mice. Computational analysis showed that skipping of exon 4 from the Tubd1 transcript results in at least 10 premature termination codons that would likely lead to transcript degradation by nonsense-mediated decay (Fig. 4H). To test this hypothesis we used emetine to block NMD in Tra2b-depleted and Tra2b-expressing murine embryonic fibroblasts. In Tra2b-depleted cells treated with emetine the Δ4 isoform of Tubd1 was strongly detectable while it was not detectable in Tra2b-expressing cells (Fig. 4J). This strongly suggests that skipping of Tubd1 exon4 results in nonsense-mediated decay of the transcript and implicates splicing regulation of that exon to be a possible mechanism of expressional regulation in vivo.

Alternatively Spliced Exons of Sgol2 and Tubd1 are Responsive to Depletion and Overexpression of HTRA2B in a Minigene Splicing Assay

We further tested splicing of the Sgol2 and Tubd1 alternative exons in a minigene splicing assay. Exons with at least 0.5 kb of flanking intronic sequence were cloned into the pSPL3 vector (Fig. 5A). Transfected into HEK293T cells the vector produces mRNA containing the genomic insert with the respective exon flaked by the tat a and tat b exons of HIV. These flanking exons are constitutively spliced resulting in a specific mRNA. Inclusion of the exon of interest results accordingly in a longer mRNA that can be detected and compared by size differences using RT-PCR (Fig. 5A). We tested the splicing of the human and murine versions of each exon while overexpressing or depleting HTRA2B using siRNAs (Fig. 5B). PCR results indicate that only the murine Sgol2 but not the human SGOL2 is responsive to HTRA2B overexpression in HEK293T cells (Fig. 5C–E). Under normal conditions (scr) Sgol2 exon 4 was included in 43% of the transcripts. After overexpression of HTRA2B exon 4 was included in 89% of the transcripts suggesting that HTRA2B effectively promotes inclusion of murine Sgol2 exon 4 (Fig. 5E). On the contrary, HTRA2B overexpression had no effect on the human SGOL2 transcripts (Fig. 5C). The human exon 4 is included in 93% of the transcripts under normal conditions (scr) (Fig. 5D). This suggests that HTRA2B overexpression is unable to further enhance SGOL2 exon 4 inclusion. As we used human embryonic kidney cells (HEK293T) for minigene splicing assays we analyzed the intrinsic splice site strengths of the tested exons by comparing to the primate splice site consensus (http://ibis.tau.ac.il/ssat/SpliceSiteFrame.htm). Indeed, scores of both the 3′ and 5′ splice sites were higher for the human sequence compared to the murine version. Furthermore, ablation of endogenous HTRA2B in HEK293T cells by RNAi did not decrease exon 4 inclusion (Fig. 5 D, E). These results suggest that the high intrinsic splice site strength of the human SGOL2 exon 4 is sufficient to constitutively splice that exon. Therefore, results obtained for Sgol2 in the mouse are likely not transferable to humans.

Download:

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

Figure 5. Splicing of Sgol2 and Tubd1 is responsive to changes in Tra2b concentration.

(A) The murine and human versions of the genomic regions comprising the identified exons of Sgol2 and Tubd1 were cloned into the pSPL3 exon trapping vector. (B) HEK293T cells were co-transfected with the pSPL3 minigene vector and siRNA specific for Tra2b or a TRA2B-GFP expression vector. Western Blot analysis shows efficiently reduced Tra2b protein levels and solid overexpression of TRA2B-GFP. (C) RNA was analyzed for exon inclusion after 48 h by semi-quantitative RT-PCR. (D–G) RT-PCR results were densitometrically quantified. Exon 4 of murine but not human Sgol2 is responsive to increased concentrations of Tra2b as splicing inclusion significantly increased from 43% to 89%. Knock-down of Tra2b is insufficient to reduce Sgol2 exon4 splicing inclusion. Exon 4 of human but not murine Tubd1 is responsive to increased concentrations of Tra2b as splicing inclusion significantly increased from 79% to 94%. Knock-down of Tra2b decreased inclusion of exon 4 from 79% to 71%. Nt, non-treated; scr, scrambled siRNA; si, siRNA against Tra2b; n.s., non significant; error bars show the s.e.m.; significance levels are *p<0.05, **p<0.01, ***p<0.001 (Student’s t-test).

https://doi.org/10.1371/journal.pone.0089020.g005

The alterantive splicing of tubulin delta chain exon 4 in the minigene is HTRA2B dose-dependent (Fig. 5C). Under normal conditions (scr) 79% of transcripts show exon 4 inclusion. Ablation of HTRA2B slightly but significantly reduced exon 4 inclusion to 71% while overexpression significantly increased inclusion to 94% (Fig. 5F). This indicates that HTRA2B promotes TUBD1 exon 4 inclusion, confirming our results obtained in the mouse model. Surprisingly, the murine version of the minigene failed to respond to altered HTRA2B concentrations (Fig. 5C). Under normal conditions (scr) exon 4 was included in 92% of the transcripts (Fig. 5G) and splicing inclusion did not change upon knock-down or overexpression of HTRA2B. This finding is in so far intriguing as we originally detected the Tubd1 exon 4 splicing event in mouse (Fig. 4 G, I). The splice site strengths of human and murine Tubd1 exon 4 are very similar and analysis suggest that the murine 3′ and 5′ splice sites are presumably equally well recognized by the human splicing machinery in HEK293T cells. However, there are minor differences in number and position of well-known binding sites (e.g. SF2/ASF and SRp40) that might be attributable to a more efficient inclusion of exon 4 in a human cellular system compared to native conditions in the mouse. Whether and to what extend splicing processes of Sgol2 and Tubd1 have a physiological impact in vivo remains to be clarified.

Cyclin Dependent Kinase Inhibitor 1a (p21) is Upregulated in Neuronal Specific Tra2b Knock-out Mice

In our previous study we identified the T-exon of the histone chaperone Nasp to be spliced by Tra2b in mouse testis and the developing brain [17]. Ablation of Tra2b decreases the inclusion rate of the Nasp T-exon resulting in a decreased expression of the tNasp isoform (Fig. 3E). Selective ablation of tNasp via RNAi has previously been demonstrated to induce apoptosis in tumor cells via expressional upregulation of the cyclin dependent kinase inhibitor 1a (Cdkn1a/p21, ENSMUSG00000023067) [35]. In agreement to this we found p21 to be upregulated by more than 1.5-fold in KO mice in our exon array analysis (Dataset S2). Using semi-quantitative RT-PCR we validated these exon array-based data. mRNA levels of p21 were significantly upregulated by 1.4-fold in a group of KO animals compared to controls or HET mice (Fig. 6A, B).

Download:

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

Figure 6. p21 is upregulated as a response to Tra2b depletion in the mouse brain and in neural stem cells.

(A,B) Semi-quantitative RT-PCR using whole brain RNA of neuronal specific Tra2b KO mice and controls. p21 is significantly upregulated by 1.5-fold in KO mice as compared to HET or control mice. p21 expression is indifferent between controls and HET mice. p21 expression was normalized to Hprt. (C–G) NSC34 neural stem cells were transfected with siRNAs specific to Tra2b or scrambled siRNAs. siRNA treatment but not scr-treatment effectively reduced Tra2b protein and mRNA levels after 24 h, 48 h and 72 h after transfection (C–E). Tra2b function was strongly reduced as the Nasp transcript showed a significantly lower inclusion of the T-exon at 24 h, 48 h and 72 hours after transfection (F). 48 hours after transfection p21 expression was found slightly but significantly increased on RNA level (G) but not on protein level (D). 72 hours after transfection p21 was massively and highly significantly upregulated on RNA and protein level by +2.2-fold (D,G). a.u., arbitrary units; nt, non-treated; scr, scrambled siRNA; si, siRNA against Tra2b; error bars show the s.e.m.; significance levels are *p<0.05, **p<0.01, ***p<0.001 (Student’s t-test).

https://doi.org/10.1371/journal.pone.0089020.g006

Due to Cre recombinase-related mosaicism in the brain of KO mice we expect that apoptotic cells will contribute only with a small fraction to the whole brain (Fig. 2A). Still we were able to detect p21 upregulation in whole brain RNA indicating a solid upregulation in the affected cell populations. To test this hypothesis we used murine neuronal stem cells (NSC34 cells) as their neuronal fate resembles the neuronal identity of affected cells found in the ventricular and subventricular cortical layers of KO mice [54]. NSC34 cells were depleted of Tra2b using specific siRNAs. RNA levels of the functional Tra2b Δ2 isoform and Tra2b protein levels were markedly reduced within 24 and 48 hours after transfection (Fig. 6C–E). As expected depletion of Tra2b caused changes in splicing of the Nasp transcript. The tNasp isoform was significantly downregulated by at least 50% within 24 to 48 hours after transfection (Fig. 6F) suggesting that Tra2b activity is functionally abolished. To test whether p21 elevation found in KO mice (Fig. 6A, B) can be reconstituted in NSC34 cells, we measured p21 expression via quantitative real-time PCR and determined p21 protein amounts via Western Blot. Within 48 hours after transfection p21 did not increase on protein level (Fig. 6C, D). However, we found a modest but significant increase of p21 expression on mRNA level (Fig. 6G). To asses long term effects of Tra2b ablation on p21 expression we assayed cells 72 and 96 hours after transfection. Notably, after 72 hours about 20% of the Tra2b-depleted cells had died (data not shown) and p21 was strongly and highly significantly upregulated by +2.2-fold on mRNA and protein level (Fig. 6C, D, G). Strikingly, after 96 hours the majority of transfected cells had died indicating a massive loss of proliferative potential and possible cell cycle arrest and apoptosis. The succession of events suggests that the downstream effects originally triggered by Tra2b ablation do not occur instantly but rather require time to establish. This is in line with our findings in KO mice, in which onset of Cre recombinase expression and thus Tra2b depletion starts at around 10.5 dpc. Abnormalities in brain development cannot be observed before 14.5 to 15.5 dpc, however. These findings indicate that Tra2b deficiency leads to p21 activation and suggest p21 activation to be a major contributor to apoptosis and developmental defects observed in neuronal-specific Tra2b KO mice.