Figure 1.
SEED genes are used to guide the selection of 8 GEO datasets. Those experiments are merged, and normalized, in order to obtain a single dataset consisting of 45101 Affymetrix 430 2 probes and 56 GSE experimental conditions. SEED genes are the our positive training examples in a positive-only Support Vector Machine (SVM) classification framework composed of 1000 classifiers. To overcome the lack of counterpart members each classifier is trained with a random subsets of genes adopted as negative examples, leaving the SEED genes fixed as positive examples. At each training/prediction run the remaining unlabeled genes are scored according to the classifier. The Main Gene Signature (MGS) hypothesis is obtained considering the top 100 genes ranked by averaging the scores among all random SVM classifiers.
Figure 2.
The overall coverage of retroelements detected with RepeatMasker (v3.3.0) was determined in a 2.5 kb window (2.0 kb upstream and 0.5 kb downstream) surrounding the 5′ terminus of the mouse genome (GRCm38/mm10) annotated transcripts. The GC percentage levels in a 2.0 kb window (1.0 kb upstream and 1.0 kb downstream) surrounding the 5′-terminus of the mouse genome (GRCm38/mm10) annotated transcripts were determined with UCSC faCount tool. The x-axes show the coverage level of retroelements or GC percentage, while the y-axes show the number of gene sets that exhibit that coverage level. The distribution of the coverage levels of 5000 randomly selected sets of genes with the same number of MGS genes is showed in grey respectively for: A) LTR (average level of random distribution 34%); B) SINE/Alu (average level of random distribution 67%); C) LINE (average level of random distribution 26%); D) GC% (average level of random distribution 51%). The vertical lines show the coverage levels of the MGS set. A significant enrichment is observed for LTR and SINE retroelements (p-value<0.001); t-test shows that the level of GC% in the promoter region of MGS genes is significantly lower (p-value = 1.27 10−15).
Figure 3.
ESCs heterogeneous subpopulation.
Gene expression pattern in ESCs cultures by in situ hybridization after 3 days in culture. A) RNAs are detected by a BCIP/NBT system producing a deep purple reaction product. RNA ISH representative “spotted” patterns on ESCs colonies (20X). The panel B shows double RNA ISH stain (red) (Dub1, Gm12794, Gm13871, and Zfp352), and Zscan4 RNA ISH staining (green), counterstained with DAPI (blue) (63x). The panel C shows double stain through Zscan4 RNA ISH (red) and NANOG immune-staining (green), counterstained with DAPI (blue) (63x).
Figure 4.
SEED and mosaic MGS expressions in ESCs.
ESCs are cultured for 5 days in: RM, Lif-, and RM supplemented with PD0325901 (0.5 µM, Stemgen), and CHIR99021 (3.0 µM, Stemgen) which prevent differentiation by inhibiting ERK, and Gsk3 (2i). The mRNA expression levels are assessed by Real Time PCR and normalized by Regular Medium (RM) condition in Lif- (panel A) and in ground state condition (panel B). The ESC differentiation condition and the ESC ground state condition are evaluated through the expression levels of key canonical markers of pluripotency such as Oct3/4, Nanog, and Rex1 that we considered as positive controls.
Figure 5.
Ablation of Gm12794 expressing ESCs.
A) Schematic diagram of the pGm12794-Strawberry reporter vector. The reporter gene Strawberry is placed under the control of a 5.0 kb region upstream the Gm12794 ATG start codon. A PGK-Neor cassette is used for the selection of the ESC clones; the Gm12794-Strawberry reporter electroporated in mESCs is visualized by the Strawberry reporter gene (20x). B) Schematic diagram of the Gm12794-HSV-Tk vector electroporated in mESCs to generate ESCGm12794_HSVTK lines. ESCGm12794_HSVTK lines and control ESCHSVTK lines were cultured in presence or absence of Ganciclovir (GCV) (2.0 µM, Sigma) (20x).