Our results show that large-scale oligonucleotide based microarrays can be used to distinguish between undifferentiated hESC and differentiating hEBs. Two sets of markers can be distinguished in a set of 'stemness" genes that are present at higher levels in undifferentiated cells whose levels are reduced as cells differentiate and a complementary set of genes that are absent or present at low levels in hESC and are upregulated as cells differentiate. The set of genes that are upregulated as cells differentiate include known as well as unknown genes. This pattern of gene expression was confirmed by a variety of independent means including microarray using a second independent array set, comparison with MPSS using similar samples and by RT-PCR. The combined set of upregulated and down regulated genes will serve as a sensitive indicator of the state of hESC and the novel sets of genes identified are likely candidates that may participate in regulating the process of differentiation.
In our previous study we demonstrated over expression of a set of 92 genes in six undifferentiated ESC cell [28
]. Expression of almost all of the genes was confirmed in the BG02-ES and pooled hESC used in the present study. The present study further refines the set of stemness genes to identify those that are rapidly down-regulated as ES cells differentiate. We show that 77 of these are down modulated when differentiated to embryoid bodies using undifferentiated BG02-ES. Further differentiation of BG02-ES to day 21 BG02-EB showed consistency in down modulation of many of these. This refined subset of previously identified stemness genes can be used to monitor the transition of undifferentiated pluripotent human embryonic stem cells to differentiated embryoid bodies. Analyzing the pattern of gene expression, we would suggest that ES specific genes such as Lin-28, PSIP2, PITX2, DNMT3B, and Galanin whose downregulation is seen in both lines and confirmed by EST scan and microarray represent good initial candidates to assess the ES cell state. Furthermore, LEFTB and CER1, inhibitors of nodal signaling, which were downmodulated as ES cells differentiated (data not shown) in addition to other members of the TGFβ signaling family may be sensitive indicators of ESC differentiation. Oct3/4, Sox-2 and nanog in contrast while expressed by ES cells may not be as good in assessing ES cell to embryoid body differentiation. While Oct-3/4 expression declined markedly in both BG02 EBs and pooled ES derived EBs compared to ES the decline was slow and not as rapid as that of other genes. This may be because Oct3/4 expression persists in germ cells that are derived from ESC [31
]. Likewise our analysis of nanog showed variable results and there was a discrepancy between the microarray and PCR results and we noted an increase in nanog levels in Day 21 BG02-EB samples by microarray. The reason for this difference is not clear. It is possible that microarray is detecting one of the eleven processed pseudogenes for nanog and that its expression on day 21 represents crosshybridization. Alternatively, nanog expression may be increased as a result of differentiation. This conclusion is supported by a recent report that nanog is expressed in mature tissues [32
]. We have confirmed that nanog expression is indeed present in some germ cells and early neuronal populations (data not shown). These tests were performed by RT-PCR using multiple primers and double label immunocytochemistry (results not shown). Future studies will examine expression of nanog in EBs using specific antibody and its function in differentiated cells. At present we recommend that nanog expression should be confirmed by PCR or immunocytochemistry and it should not be included as an initial indicator of the ES cell state and differentiation.
In addition to confirming the downregulation of hESC specific genes our study identified a set 333 genes that were uniquely overexpressed at ≥ 3 fold in day 13 BG02 EB compared to undifferentiated BG02 ES cell lines. Out of these 333 genes 194 genes also showed ≥ 3 fold over expression in EB samples prepared from WiCell lines. This significant similarity of unique genes in BG02 and WiCell line derived EBs suggest that this subset of genes can be classified as EB specific (induced as ESC differentiate) and that embryoid bodies derived from different ES cells maintain significant similarity with each other. These EB specific genes like the ES cell specific genes may be useful in distinguishing EBs from ES cells.
A detailed analysis of upregulated genes in EB indicated that a) differentiation genes, b) cell signaling, cellular process and cell cycle genes, c) cytoskeleton or cell motility genes and d) metabolism and DNA and RNA related genes were modulated. In addition, 27 hypothetical genes, 8 unknown genes and 2 zinc finger genes were upregulated in BG02 derived EB and WiCell lines derived EB. The status of upregulated genes identified by microarray studies was confirmed by MPSS analysis. Among 194 genes, 148 genes showed higher expression in EB by MPSS analysis. Overexpression of 33 genes in EBs compared to HuURNA was confirmed using alternate large-scale array. The expression of a subset of the remaining 13 genes was confirmed by RT-PCR analysis. These results suggest that microarray analysis is a useful complement to MPSS and may identify a overlapping but non-identical set of genes. Among the genes that were upregulated in both EB samples is HAND1 (heart and neural crest derivatives 1). HAND1 was dramatically up regulated (40 fold compared to 6 fold in undifferentiated ES cells). HAND1 belongs to basic helix-loop-helix family of transcription factor, may be required for early trophoblast differentiation as well as the differentiation of heart tissue [33
Additionally we noted that several markers of early differentiation such as keratin, actin and beta tubulin, which are present at low levels in ESC, were dramatically upregulated as EBs differentiated. We suggest that these markers can serve as sensitive indicators of the process of differentiation and when coupled with the down regulation of ESC markers may reliably distinguish the state of ES cell cultures. Future experiments will assess the sensitivity of a combination of markers identified in this study to temporally profile the process of differentiation. A recent study (34) has compared gene expression of 24-marker genes between EBs differentiated with eight different growth factors and hESC. Similar to our observation by microarray analysis, this study showed over expression of genes related to ectoderm differentiation such as Keratin-8, Keratin-18 and Keratin-19 when differentiated by retinoic acid. A highly significant upregulation of mesodermal related genes e.g., HAND1 was also observed which confirms our observation.
Our results further show that many genes that are induced or upregulated as EBs differentiate remain elevated at Day 21 of differentiation. These genes represent good markers for distinguishing differentiated cells from ESC. In addition we noted that several genes showed differential expression when levels of expression in Day 13 and Day 21 BG02-EB's was compared. Thirty-three of the 194 genes upregulated at Day 13 were down modulated at day21 EB. Other genes such as NPHP3, CAPN1, and CNTN6 were up regulated in day 21 BG02-EB's when compared to day 13 EB. These later appearing genes may reflect additional derivatives that appear as EBs undergo further differentiation. These later appearing markers along with markers that are down regulated can be used to distinguish early from late EBs.
A recent study has characterized differences in gene expression between mouse and human ESC (35). This study identified differences between mouse and human ESC and reported that differences were species specific rather than arising from differing culture conditions. In our current study, while the expression of many genes was similar in rodent and human cells significant differences were found. For example the expression of vimentin, eomesodermin, SSEA4, AFP, IL6ST and HEB was found in hESC but not in mouse ES cells.
The large oligonucleotide arrays have allowed us to build a comprehensive data set that includes hundreds of genes (both novel and unknown) that are either upregulated or downregulated as hESC differentiate. The list of genes has been validated by testing on different cells lines, testing using different oligonucleotide sequences to different regions of the genes and testing arrays used in different formats. Our results show clearly that while small differences exist as different techniques are used the core set of markers is quite robust, the differences in expression quite large and are readily detectable even using a relatively insensitive method such as a microarray. Our results show further that restricting the analysis to only this set of genes as would be necessary in focused array does not reduce the sensitivity of the assessment. Indeed, based on our results we would suggest that an order of magnitude fewer genes may be sufficient if they are appropriately selected from the lists. We have recently developed a focused array that includes genes that are down regulated as hESC differentiate as well as those that are induced as cells differentiate [36