Using expression data from teratomas we
Using expression data from teratomas, we were able to identify a large variety of differentiated tissues within them. This allowed us to analyze a repertoire of differentiated cell types in each embryonic germ layer. Teratomas initiated from PSCs express characteristic genes of all analyzed tissues and a similar expression of all lineages. Other tumors are typically composed of uniform and homogenous tissues, as seen by our 100-gene scorecard. We show that tumors originating from all lineages as well as from extraembryonic tissues express the respective genes in an exclusive manner. Weighing this expression creates a score that successfully separates PSC-derived mature tumors from tumors originating from other cell types. While this separation is evident, it is unclear why some teratomas provide a much higher grade in the TeratoScore analysis. We were not able to find a common ground for high or low scores, looking at karyotype integrity or genetic identity. In our study, we have analyzed two sets of different teratomas generated from the same cell types (CSES22 and CSES45; see Table S1). The TeratoScore of both pairs of teratomas was higher than 100, affirming the cell lines’ pluripotency, but the numbers were somewhat different (160 and 315 for CSES22-initiated tumors and 234 and 429 CSES45-initiated tumors). Examining a much larger teratoma sample from various sources and passages, including partially reprogrammed cells, might aid in identifying a biological origin for TeratoScore performance variation. It is possible that some of the variance originates from TeratoScore’s limited sensitivity. Although we show that a 100-gene scorecard is enough to discover the existence or absence of tissues and lineage, it is insufficient in identifying possible lineage differentiation biases that might affect the TeratoScore outcome. This is also true for the measurement of pluripotency markers’ expression. Pluripotent foci were previously shown to exist in teratomas initiated from trisomy 12 AICAR (Ben-David et al., 2014). While a small subset of pluripotent markers could not detect residual undifferentiated PSCs, analysis of the top 200 genes expressed in hPSCs revealed that 25 of them were upregulated T12 teratomas compared with WT teratomas (p < 0.06, following Bonferroni correction for multiple testing) compared with only 5 in T21 teratomas (p < 0.8). Several studies have analyzed the differentiation propensity of hPSCs in vitro, suggesting that different cell lines have differentiation bias, resulting in unequal ability to differentiate toward specific cell types (Bock et al., 2011; Boulting et al., 2011; Osafune et al., 2008). A method to evaluate these biases has been proposed (Bock et al., 2011), and their origins have been suggested to be both genetic and epigenetic (reviewed in Cahan and Daley, 2013). Furthermore, the International Stem Cell Initiative (ISCI) (Andrews et al., 2005) is currently conducting a large-scale study to compare different methods to analyze the pluripotency of human stem cells. Our aim in generating TeratoScore was to give a quantitative measure to differentiation of PSCs in vivo as a way to evaluate the pluripotency of the cells. Teratomas have long been argued to be an intriguing developmental model; however, the extent of their resemblance to typical development is under debate (Ozolek and Castro, 2011). Using karyotypically abnormal cells, we were able to show significant changes in tissue-enriched gene expression. Although these changes were not very significant in the 100-gene scorecard made to identify pluripotent origins, they were highly significant when looking at the top 200 genes expressed in each tissue. Additional copies of chromosomes 12 or 21 caused a substantial increase in expression of genes related to the CNS. This upregulation may indicate these aneuploidies influence in vivo, for it is known that both conditions cause substantial intellectual disabilities (Chapman and Hesketh, 2000; Segel et al., 2006). Furthermore, case studies have reported joint and muscle malfunction in patients with Pallister-Killian mosaic syndrome (Speleman et al., 1991)—another significant trait in our analysis. It is important to note that we could not find a report of patients bearing a complete trisomy of chromosome 12. We can cautiously assume that this aberration is lethal—perhaps due to extant of developmental abnormalities implied in our analysis. In contrast, trisomy 21 seems to cause a significant downregulation in skeletal muscle-specific genes only—perhaps correlating with the successful birth of Down syndrome patients and their typical muscle hypotonia (Cisterna et al., 2013).