Undifferentiated hESCs express only a subset of TFIID TAFs
In a search of published expression datasets (Abeyta et al., 2004
), we found that several TAFs of the canonical TFIID complex were apparently not expressed in hESCs. To investigate this possibility, we analyzed expression of 13 TAFs by immunoblotting lysates from H9 cells, a well-characterized hESC line. As a control, we also analyzed TAF expression in HeLa cells, which have been extensively used to study TFIID composition and function. The immunoblot of shows, as expected, that all 13 TAFs were expressed in HeLa cells. By contrast, hESCs clearly expressed TAFs 2, 3, 5, 6, 7 and 11, whereas expression of TAFs 1, 4, 8, 9, 10, 12, and 13 was undetectable. Interestingly, TAF6 is expressed in both cell types, but the isoform present in H9 cells is predominantly the short delta form, whereas in HeLa cells, the major TAF6 isoform is the larger, alpha/beta form. The specificity of each TAF antibody was confirmed by RNA interference (RNAi)-mediated knockdown (). We observed a similar TAF expression pattern in a second hESC line, H1 cells (). Quantitative RT-PCR (qRT-PCR) analysis comparing TAF
mRNA levels in HeLa and H9 cells correlated with the immunoblotting results (). Unlike the TAFs, all other GTFs analyzed were comparably expressed in HeLa and H9 cells (). Based upon these results we conclude that only six of the canonical TFIID TAFs are present in hESCs.
Undifferentiated hESCs express only a subset of TFIID TAFs.
We next asked whether differentiation of hESCs results in a change in TAF composition. Toward this end, H9 cells were treated with retinoic acid to induce differentiation and TAF expression was analyzed by immunoblotting. shows, as expected, that following retinoic acid treatment, expression of the pluripotency factor OCT4 was lost and NES, a neuroectoderm marker, was induced. Significantly, TAFs 1, 4, 8, 9, 10, 12, and 13, which are not expressed in undifferentiated H9 cells, were induced following retinoic acid treatment. TAFs 2, 3, 5, 6, 7 and 11, which are expressed in undifferentiated H9 cells, were also present at a relatively constant level following retinoic acid treatment.
hESCs have a non-canonical TBP-containing TAF complex
To investigate whether the six hESC TAFs were associated in a stable complex, H9 cell nuclear extract was fractionated by sucrose gradient sedimentation and individual fractions analyzed for TAFs 2, 3, 5, 6, 7 and 11 by immunoblotting. The results of show that TAFs 2, 6, 7 and 11 co-sedimented with an apparent native molecular mass of ~440 kDa. By contrast, TAFs 3 and 5 fractionated heterogeneously, and a substantial portion of both TAFs had an apparent molecular mass consistent with that of the free proteins (~140 and ~100 kDa, respectively). As expected, TBP, which is associated with multiple complexes involved in transcription by all three RNA polymerases, fractionated heterogeneously. Notably, however, a peak of TBP co-sedimented with TAFs 2, 6, 7 and 11.
hESCs have a non-canonical TBP-containing TAF complex.
To provide additional evidence for a stable, multi-subunit TAF complex, and to determine whether TBP was a component, we performed co-immunoprecipitation experiments. TBP was immunoprecipitated from H9 cell nuclear extracts, and the immunoprecipitate was analyzed by immunoblotting for the six hESC TAFs. The results of show that TBP was stably associated with TAFs 2, 6, 7 and 11 but not TAFs 3 and 5. Collectively, these results indicate that H9 cells contain a non-canonical TAF complex composed of TBP and TAFs 2, 6, 7 and 11 but not TAFs 3 and 5.
Two classes of active hESC genes based on TAF promoter occupancy
The results of the expression analysis and biochemical experiments implied that the PIC formed on the promoters of active genes in hESCs would have an atypical TAF composition. To investigate this issue, we performed a series of chromatin immunoprecipitation (ChIP) experiments. In the first set of experiments we selected 10 transcriptionally active genes and performed ChIP analysis to monitor promoter occupancy by the six hESC TAFs. As a normalization standard, we also monitored occupancy of TBP and RNA polymerase II large subunit (POL2) on these 10 promoters. As expected, we found that TBP and POL2 were present at comparable levels at each of the 10 promoters (). However, the absolute level of TBP and POL2 bound to each promoter significantly varied among the 10 genes. Therefore, in this experiment and those presented below, TAF recruitment was normalized to the level of TBP occupancy. The ChIP results of revealed two groups of genes with distinct TAF promoter occupancy patterns. The first group, which we refer to as class I genes, were bound by TAFs 3 and 5 but not by TAFs 2, 6, 7 and 11, whereas the second group, class II genes, were bound by all six hESC TAFs.
Two classes of hESC genes based on TAF promoter occupancy.
To support this conclusion, we also compared TAF occupancy in HeLa and H9 cells across seven class I genes that are transcriptionally active in both cell types (). As a control, we first analyzed a representative subset of TAFs that are expressed in HeLa but not H9 cells. The ChIP analysis of shows that TAFs 1, 8 and 9 were readily detected on the promoters of genes in HeLa cells but, as expected, not H9 cells. Next, we analyzed the six TAFs that are expressed in both HeLa and hESCs. shows that in both cell types TAFs 3 and 5 were recruited to the promoters of the seven genes. By contrast we found that TAFs 2, 6, 7 and 11 were bound to the promoters of the seven genes in HeLa but not in H9 cells (). These results indicate that on the same transcriptionally active gene the TAF composition is strikingly different in HeLa and H9 cells.
Comparison of TAF promoter occupancy on an identical set of transcriptionally active genes in HeLa and H9 cells.
We next performed a series of RNAi experiments to determine the relationship between TAF occupancy and transcriptional activity. The qRT-PCR results of show that siRNA-mediated knockdown of TAFs 3 and 5 in H9 cells () greatly reduced expression of both class I and II genes. By contrast, siRNA-mediated knockdown of TAFs 2, 6, 7 and 11 decreased expression of class II, but did not affect expression of class I genes (). Comparable results were obtained with a second, unrelated siRNA directed against each of the six TAFs (). Collectively, these results establish a strong relationship between TAF occupancy and transcriptional activity in hESCs.
A strong relationship between TAF occupancy and transcriptional activity in hESCs.
Global ChIP-chip analysis of TAF occupancy in hESCs
To confirm and extend the ChIP results, we performed global ChIP-chip analyses. In these experiments we monitored, in parallel, promoter occupancy of TAFs 3 and 5, TAFs 7 and 11 (as representative members of the TBP/TAF 2,6,7,11 complex), TBP and POL2. The overall results are summarized in . We first defined a group of ~3600 high-confidence actively transcribed genes based upon co-occupancy of both TBP and POL2 at the transcription start-site (). The vast majority of active genes had promoter-bound TAF3 and TAF5 (). Significantly, a smaller fraction of active genes had promoter-bound TAF7 or TAF11, and there was substantial overlap between TAF7- and TAF11-bound genes (). As expected, the vast majority of genes bound by TAFs 7 and 11 were also bound by TAFs 3 and 5 (). Representative examples of promoter occupancy maps for two class I (SLC25A3, CCNB2) and class II (SFPQ, UCHL1) genes are shown in .
Global ChIP-chip analysis of TAF occupancy.
To validate the ChIP-chip results, we selected a representative set of 44 genes and performed directed ChIP experiments using promoter-specific primer pairs. These validation experiments, which are shown in , confirmed the predicted TAF occupancy patterns for ≥85% of the genes tested. For example, consistent with the ChIP-chip results, there was no significant binding (i.e., no enrichment relative to the no antibody negative control) of TBP or TAFs 3, 5, 7 or 11 to a group of inactive promoters predicted by the ChIP-chip analyses to not be bound by these factors (). Most importantly, shows that TAF7 occupancy validated at 24 of 27 predicted sites (88.9%), TAF11 occupancy at 27 of 28 predicted sites (96.4%), TAF3 occupancy at 36 of 39 predicted sites (92.3%), and TAF5 occupancy at 39 of 39 predicted sites (100.0%).
Validation of ChIP-chip results by directed ChIP experiments using promoter-specific primer pairs.
also shows that the overlap between TAF7- and TAF11-bound genes was higher than that predicted by the ChIP-chip analyses. Specifically, the results show that TAF7 was present at 7 of 7 predicted TAF11-only sites, and TAF11 was present at 5 of 6 predicted TAF7-only sites, indicating that TAF7 and 11 co-occupancy was 12 of 13 (92.3%).
Moreover, as expected, genes bound by TAFs 7 and 11 were also co-occupied by TAFs 2 and 6 (). For example, of the 16 promoters analyzed that were bound by TAF7 and TAF11, TAF2 was present at 16 and TAF6 was present at 15 of these promoters. Collectively, the ChIP-chip analyses, in conjunction with the results described above, confirm the existence of two groups of genes in hESCs whose promoters are bound either only by TAFs 3 and 5 (class I genes) or by all six hESC TAFs (class II genes).
Finally, we analyzed the ChIP-chip dataset in relation to previous genome-wide studies in hESCs (Abeyta et al., 2004
; Boyer et al., 2005
; Guenther et al., 2007
) for features that might distinguish class I and class II genes and found two statistically significant differences (; ). First, the promoters of class I genes had greater histone H3 lysine 4 trimethylation (H3K4me3) than those of class II genes (). Second, the fraction of genes with alternative promoters (identified based upon UCSC Genome Browser annotations; see ‘Materials and methods’) was significantly higher for class II than for class I genes (). Representative examples of promoter occupancy maps for two class II genes with alternative promoters are shown in .
Statistical analysis of ChIP-chip data
The composition of hESC TAFs is required for proper regulation of gene expression and maintenance of the undifferentiated state
Finally, we analyzed whether the unusual composition of TAFs was important for the characteristic ability of hESCs to maintain an undifferentiated state and self-renew. shows that shRNA-mediated knockdown of each of the six hESC TAFs () induced differentiation, as evidenced by a decreased percentage of alkaline phosphatase-positive colonies. To confirm this conclusion, we also tested whether knocking down hESC TAFs would induce differentiation by analyzing expression of a diverse set of differentiation markers: AFP (endoderm), CGB7 (trophoectoderm), IGF2 (mesoderm), NES (ectoderm) and SOX1 (neuroectoderm). shows that depletion of each hESC TAF resulted in up-regulation, to varying extents, of these differentiation markers. Comparable results were obtained with a second, unrelated shRNA or siRNA directed against each of the six TAFs (). Finally, the induction of differentiation following knockdown of hESC TAFs was also evidenced by decreased expression of the pluripotency genes NANOG ( and ) and OCT4 ( and ). Thus, the hESC TAFs are required to maintain the undifferentiated state.
The composition of hESC TAFs is required for maintenance of the undifferentiated state.
In a reciprocal set of experiments, we altered TAF composition by ectopically expressing a TAF that is not normally present in hESCs. We found that ectopic expression of TAF1 () resulted in differentiation, as evidenced by a decreased number of alkaline phosphatase-positive colonies (), the induction of differentiation markers (, left), and decreased expression of NANOG (, right). Interestingly, shows that ectopic TAF1 expression decreased expression of several class II genes, including as expected OCT4, whereas expression of class I genes was either unaffected or in some instances increased modestly. Ectopic expression of several other TAFs not present in hESCs also resulted in loss of expression of OCT4, a class II gene, but not ACTB, a class I gene (). By contrast, ectopic expression of TBP or TAFs that are present in hESCs did not affect OCT4 levels. To investigate the basis for the decreased OCT4 expression, we performed ChIP analysis. shows that, following ectopic expression of TAF1, TBP and POL2 were no longer recruited to the OCT4 promoter. Collectively, the results of show that altering the composition of hESC TAFs results in misregulation of gene expression and induction of differentiation.