On Human Promoters

Global distribution of negative cofactor 2 subunit-␣ on human promoters

Thomas K. Albert*, Korbinian Grote†, Stefan Boeing*, Gertraud Stelzer*, Aloys Schepers‡, and Michael Meisterernst*§

Departments of *Gene Expression and ‡Gene Vectors, GSF–National Research Center for Environment and Health, Marchioninistrasse 25, 81377 Munich, Germany; and †Genomatix Software GmbH, Bayerstrasse 85a, 80335 Munich, Germany

Communicated by Robert G. Roeder, The Rockefeller University, New York, NY, April 30, 2007 (received for review November 16, 2006) Negative cofactor 2 (NC2) forms a stable complex with TATA- of multiple sequence-specific transcription factors in mammalian binding protein (TBP) on promoters in vitro. Its association with TBP genomes. However, surprisingly little is known about the genome- prevents the binding of TFIIB and leads to inhibition of preinitiation wide location of GTFs and general cofactors in mammalian cells. complex formation. Here, we investigate the association of NC2 One exception is the investigation of the genome-wide location of subunit-␣ with human RNA polymerase II promoter regions by TAF1 conducted along with RNAPII (16). using gene-specific ChIP and genome-wide promoter ChIPchip In our analysis, the human cofactor NC2␣ occupied Ͼ20% of analyses. We find NC2␣ associated with a large number of human all human gene promoters. We observed a positive correlation promoters, where it peaks close to the core regions. NC2 occupancy of NC2 gene occupancy with mRNA levels. On the other hand, in vivo positively correlates with mRNA levels, which perhaps NC2 occupancy negatively correlated with the presence of BRE, reflects its capacity to stabilize TBP on promoter regions. In single the TFIIB core promoter recognition element. Collectively, the gene analyses, we confirm core promoter binding and in addition data support the notion of a global role for NC2 in gene control. map the NC2 complex to enhancer proximal regions. High- At the single gene level the NC2/TFIIB ratio is controlled in cis. occupancy histone genes display a stable NC2/TFIIB ratio during the cell cycle, which otherwise varies markedly from one gene to Results another. The latter is at least in part explained by an observed Genome-Wide Promoter Association of NC2␣. ChIPchip analysis was negative correlation of NC2 occupancy with the presence of the conducted with high-density promoter tiling arrays that cover TFIIB recognition element in core promoter regions. Our data Ϸ24,000 human genes. NC2-DNA complexes were precipitated establish the genome-wide basis for general and gene-specific from a human Epstein–Barr virus (EBV)-transformed lympho- functions of NC2 in mammalian cells. blastoid B cell line, LCL721, by using a rat monoclonal antibody directed against the ␣-subunit of the NC2 complex (17). Pro- genome-wide ͉ occupancy moters were subsequently ranked according to the median of the fold enrichments (in comparison with input DNA) on 15 probes ranscription control of gene expression in eukaryotes in- for each individual promoter. Of the 24,099 promoter regions ␣ volves a dynamic interplay of positively and negatively acting analyzed, 2,759 (11.4%) displayed high NC2 levels with a fold T Ͼ factors. The central nucleation factor for the assembly of RNA enrichment of 1.4, and another 3,517 promoters (14.6%) ␣ Ͼ polymerase II (RNAPII) initiation complexes at promoters is showed medium NC2 enrichment of 1.2 (Fig. 1b). The TBP, the TATA-binding protein (1). Binding of TBP to the remaining 17,823 genes displayed fold enrichments below the ␣ TATA box is a rate-limiting step during preinitiation complex threshold of 1.2 and represent genes with low or no NC2 (PIC) formation. Multiple factors target TBP or contribute to promoter occupancy [supporting information (SI) Data Set 1]. the formation of TBP–DNA complexes by using distinct mech- As a control, ChIPchip was conducted in parallel with a anisms. Specific examples include general transcription factors nonspecific IgG monoclonal antibody (SI Fig. 7). The alignment ␣ (GTFs) such as TFIIA and TFIIB, TBP-associated factors of both data sets indicated high specificity of the NC2 ChIP. (TAFs), and a group of proteins termed cofactors, among them Only a small fraction of the promoters (269/6,089, or 4.4%) with Ͼ ␣ BTAF1/Mot1p and negative cofactor NC2 (2). a median fold enrichment 1.2 in the NC2 ChIP showed a NC2 was originally defined as a TBP-binding activity in human comparable enrichment in the control ChIP (SI Fig. 7). nuclear extracts that repressed RNAPII transcription (3). The Although we clearly identified peaks on numerous genes with Ͻ cofactor is composed of two subunits, NC2␣ and NC2␤, that median fold enrichment 1.2 (for selected examples see SI Fig. dimerize by means of histone-fold domains. The NC2 complex 8), the choice of the thresholds remains arbitrary. To substan- binds to TBP–DNA from the underside, thereby preventing the tiate their relevance, we plotted the complete set of data against binding of TFIIB (Fig. 1a). Biochemical and genetic data the annotated main start sites of the genes. The group of suggested that NC2 further interferes with recruitment of promoters representing median fold enrichment of 1.2 to 1.4 TFIIA, thereby counteracting PIC assembly (3–7). showed maximum probe signals near the transcription start site The genes encoding NC2 are absent in prokaryotes including (Fig. 1c). Peak formation is even more pronounced in the group of promoters with median fold enrichment Ͼ1.4. In contrast, Archea but are conserved in all eukaryotes. In yeast, the corre- ␣ sponding genes, BUR6 and YDR1, emerged as transcriptional low-occupancy NC2 target genes and control sets showed a cofactors in independent genetic screens (8–10). Yeast gene expression profiling confirmed repressive functions but also Author contributions: T.K.A. and M.M. designed research; T.K.A., S.B. and G.S. performed revealed stimulatory effects of NC2 on gene transcription (11– research; A.S. contributed analytical tools; T.K.A. and K.G. analyzed data; and T.K.A. and 13). The cofactor associates with yeast and human promoters in M.M. wrote the paper. vivo. Mammalian NC2 displayed little specificity for TATA. The authors declare no conflict of interest. Consistently, both TATA and TATA-less core promoter were Abbreviations: ChIPchip, ChIP on DNA microarrays; NC2, negative cofactor 2; RNAPII, RNA recognized by NC2 in vivo (11, 14). polymerase II; TBP, TATA-binding protein; TCR␤, T cell receptor ␤. Here, we initiated chromatin immunoprecipitation on microar- §To whom correspondence should be addressed. E-mail: [email protected]. rays (ChIPchip) investigations with human NC2 in B cells. First This article contains supporting information online at www.pnas.org/cgi/content/full/ applied to low-complexity genomes such as those of yeast (15), 0703490104/DC1. ChIPchip has now been successfully used to determine the location © 2007 by The National Academy of Sciences of the USA

10000–10005 ͉ PNAS ͉ June 12, 2007 ͉ vol. 104 ͉ no. 24 www.pnas.org͞cgi͞doi͞10.1073͞pnas.0703490104 Downloaded by guest on October 2, 2021 a TBP TBP

TATA TATA

NC2α

NC2β TFIIB

b 3% 4%

9% 16% 28% 15% NC2α Control

46% 80%

Median fold N Occupancy enrichment Fig. 2. NC2␣ ChIPchip proﬁle on individual promoters. Shown are NC2␣ occu- 616 Very high > 2.0 pancies at the selected ribosomal protein gene RPL7A (Left) and at the RNA 2,143 High 2.0 - 1.4 processing factor gene CSTF3 (Right) in LCL721 cells. Log2 ratios of the NC2␣ ChIP 3,517 Medium 1.4 - 1.2 versus input DNA on individual array probes are plotted. The approximate width 11,008 Low 1.2 - 1.0 6,815 No < 1.0 of the peak area is outlined. The broken arrows mark transcription start sites.

80 c )%( the mean of all signals 7,154 peaks were detected (low-

s 60 confidence set). In the corresponding set of data set 2, 72.5% of e b

o the peaks identified under stringent conditions in data set 1 are r p f p 40 NC2α cut-off >1.4 present . If the top 5% or 10% of median fold enrichments in o o

n NC2α cut-off >1.2 data set 2 were compared with the top 25% of data set 1 o it 20 α Ͼ c NC2 no cut-off (referring to all genes with a median fold 1.2), 92% and 83%, arF Control respectively, of the genes were recovered. Similar numbers were 0 0 004 0 0021- 0 obtained with alternative algorithms (18). As documented below 0 0 4 8 - - (Fig. 3) the fraction of NC2-positive genes further increases if Distance from TSS (bp) only active genes are taken into account. Furthermore, our own unpublished observation show that occupancy numbers of NC2␣ ␣ Fig. 1. Genome-wide promoter association of NC2 .(a) Structure of the are close to the one observed for the general factor TFIIB NC2–TBP–TATA complex (31) and of the TFIIB–TBP–TATA complex (38). (b) Charts showing the distribution of enrichment classes of 24,099 human pro- (T.K.A. and M.M., unpublished observation). Altogether, these moters in a NC2␣ ChIP (Left) and an isotype IgG control ChIP (Right) performed data support the argument for global gene occupancy and for a in LCL721 B cells. Five enrichment classes were ranked according to the general role of NC2 in gene transcription. indicated median fold enrichments on 15 probes per individual promoter. (c) Distribution of the distance between NC2␣-bound probes and the closest NC2␣ Occupancy Positively Correlates with mRNA Levels. Affymetrix transcription start site (TSS). The fraction of bound probes that map to 100-bp microarrays (SI Data Set 2) were used to correlate NC2␣ BIOCHEMISTRY intervals from the TSS at the indicated cut-offs is shown. The averaged distribution of the nonthresholded NC2␣ (no cut-off) and control ChIPchip set occupancy with gene expression in LCL721 B cells. The analysis is also depicted.

10,000 stochastic distribution (Fig. 1c). The inspection of the individual genes confirms the statistical analysis at the single gene level. As 1,000 is shown here for the RPL7A gene, the experimental peak width was often in the range of 300–400 bp, which is near the mean

experimental resolution (Fig. 2 Left). Somewhat broader areas of n 100 expression oi

occupancy (600–800 bp), but still well below the range moni- sse r

tored on the array, were observed for several genes, as exem- p

x 10 plified here by CSTF3 (Fig. 2 Right). E A second biological ChIPchip replicate was conducted. The second data set yielded on average higher fold enrichments (26% 1 of the genes showed a value Ͼ2.0). In our experience, these variations are generally observed unless the experiments are 0.1 conducted in parallel throughout the entire ChIPchip procedure. 0.3 110 Nevertheless, the second data set confirms the observation of Median fold enrichment occupancy of a significant percentage of mammalian genes. The Fig. 3. Positive correlation of NC2␣ occupancy and gene expression. Scatter latter data set was analyzed by statistical means by using peak plot depicting NC2␣ occupancy (median fold enrichment of NimbleGen probe finding algorithm M-Peak (16). Under stringent conditions sets, x axis) and corresponding gene expression (Affymetrix probe set signals, (peak threshold was the mean of all signals plus 2.5 times the y axis) in LCL721 cells. A moving average of NC2␣ enrichment on genes calculated standard deviation, P ϭ 0.05), 3,241 peaks (high- (window size 0.5, step size 0.1) is indicated by black dots. The horizontal line confidence set) were identified. If the threshold was reduced to represents the mean of signals on all Affymetrix probe sets.

Albert et al. PNAS ͉ June 12, 2007 ͉ vol. 104 ͉ no. 24 ͉ 10001 Downloaded by guest on October 2, 2021 β β a kb 0.5 11.5 c G1 a V C1 MAR E

)2gol 1 Control 0 -1000 +1 +1000 1 kb ( tne ( H2BJ H2AG 3 S 1824635 7 9101112 13 14

mhcirnE 2 NC2α

1 evitaleR ycnapucco 1.0 b NC2α 0 G2/M 0.6 b kb 0.5 1 1.5 H2BJ H2AG 0.2

1 2468 y 1.0 S 35 7 d c NC2α 1234567891011121314 na G1 tu 3 p G2/M u

pni tnecreP pni 0.6 ycnapucco evitaleR ycnapucco cc 1.0 2 c NC2β o o . l 0.2 1 eR 0.6 0 12345678 12345678 0.2 ycnap 1.0 y 1.0 e TFIIB c TBP napucco napucco 1234567891011121314

ucco .leR ucco 0.6 0.6 Fig. 5. NC2 binding to TCR␤.(a) Scheme of the rearranged TCR␤ locus in . 0.2 l 0.2 Jurkat cells. The functional transcription unit extends from the variable region e

R V␤ to constant region C1. Distal downstream elements such as matrix attach- 12345678 12345678 ment region (MAR) and the 3Ј enhancer (E␤) are also shown. The numbered gray boxes indicate the qPCR amplicons used to monitor binding of NC2␣ (b) Fig. 4. NC2 occupancy of histone genes. (a) ChIPchip proﬁle of NC2␣ and NC2␤ (c) in Jurkat T cells. occupancy at the histone H2BJ-H2AG locus (log2 scale; y axis). For comparison, the binding proﬁle from the control ChIP is shown above the bar graph. (b) ChIPchip validation of NC2 binding to H2BJ-H2AG in Jurkat cells by qPCR. Relative positions of amplicons 1–8 are shown above the bar graph. (c) FACS ment in the center of the promoter region, whereas the control analysis of elutriated Raji cell fractions in G1, S, and G2/M phase. (d and e) ChIP displayed background levels throughout the locus (Fig. 4a). Conventional ChIP analysis of relative NC2␣ (d) and TFIIB and TBP (e) occu- Eight primer pairs were selected for specific priming on this pancy levels at H2BJ-H2AG in phased Raji cells. histone locus. NC2 occupancy peaked at the bidirectional promoter and dropped to low levels in the H2A and H2B coding regions (Fig. 4b), thereby reproducing the ChIPchip result in the suggested an overall positive correlation of mRNA levels with peak area. Beyond it, the array data displayed peak shoulders in NC2␣ occupancy (Fig. 3). Seventy-three percent of all genes ␣ the coding region, which are absent in the PCR analysis (Fig. 4 showing very high NC2 occupancy (i.e., above a median 2-fold a vs. b). This is perhaps explained by cross-hybridization of the enrichment) were expressed to above-average levels, whereas ␣ multiple gene copies of the highly conserved histone genes. This this number dropped to 21% for genes without NC2 .Onthe analysis provides an independent confirmation of the ChIPchip other hand, there was no strict requirement for the presence of ␣ results. NC2 on genes that produce high levels of mRNA. For example, The expression of histone genes is regulated during the cell NC2␣ was apparently absent in a group of 768 genes whose cycle (20). We next asked whether the changes in expression mRNA levels were above average (SI Data Set 3). Genes in the relate to NC2 occupancy. Raji lymphoma B cells were synchro- latter group included, e.g., ␤-actin and ␣-tubulin (Affymetrix nized by elutriation centrifugation to yield homogenous popu- signals 8,887 and 6,507; SI Fig. 9). Moreover, a significant lations of G ,S,andG/M phase cells as documented by FACS number of genes that had low mRNA levels showed prominent 1 2 ␣ Ͼ (Fig. 4c). The binding profile from phased Raji cells revealed a NC2 binding (median fold enrichment 1.4) within the core ␣ promoter. Examples for this are TAF5 and GCET2 (SI Fig. 9). sharp peak of NC2 at the bidirectional promoter, which was Functional gene ontology categorization revealed no signifi- maximal in S-phase and roughly 2-fold lower in G1 and G2/M cant enrichment of distinct gene families within either group (Fig. 4d). A subsequent ChIP analysis of TBP and of TFIIB (data not shown). Instead, cross-referencing of the complete binding to the histone locus revealed a similar picture: corre- NC2␣ ChIP data set with a resource of 575 human housekeeping lating to an approximately 2-fold increase of histone mRNA genes revealed prevalence of the latter (19). Furthermore, the levels in S-phase (SI Fig. 12a), the concentration of all three strongly expressed B cell-specific surface markers are repre- factors increased on the locus (Fig. 4 d and e). A Western blot sented in the group of high-occupancy genes (SI Fig. 10). analysis conducted in parallel revealed unchanged protein levels ␣ Prominent examples of the first group are histone and ribosomal of NC2 , TFIIB, and TBP in the corresponding whole-cell protein (RP) genes: 78% (61 of 78) of histone promoters and lysates (SI Fig. 12b). These ChIP data corroborate earlier 71% (51 of 72) of RP promoters carry high levels of NC2 findings that showed stable association of TBP, TFIIB, and (median Ͼ1.4, SI Fig. 11). Conversely, NC2 is absent in tissue- NC2␤ at histone promoters in mitotic chromatin of HeLa cells specific genes that are not expressed in B cells, encoding, for (21). NC2 binds histone promoters throughout the cell cycle, in example, the receptors for surface markers in hematopoietic cell which NC2␣ and TFIIB occupancy ratios remained roughly lineages such as T cells or natural killer cells (data not shown). constant. Notably, close inspection revealed that NC2/TFIIB ratios differed between H2BJ (amplicon 3) and H2AG (ampli- Specificity of NC2␣ ChIP Documented by Single Gene Analysis. To cons 4 and 5 in Fig. 4). verify the localization of NC2 specifically at highly active genes To gain insight into binding of NC2 outside of promoter we chose to investigate a representative histone gene pair regions, we next studied a single large gene locus. For historical H2BJ–H2AG that is under control of a bidirectional promoter reasons, we chose the T cell receptor ␤ (TCR␤) gene locus region using quantitative single gene ChIP (Fig. 4). In ChIPchip, comprising the rearranged V␤ 12-3 promoter some 22 kb the bidirectional H2BJ–H2AG locus showed a Ͼ8-fold enrich- upstream of a distal 3Ј enhancer (E␤) (Fig. 5a). Fourteen pairs

10002 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0703490104 Albert et al. Downloaded by guest on October 2, 2021 Occupancy median fold enrichment a b 1.2 2.5 no/low occupancy medium / high occupancy 75 MTE 2.0 tupnI tnec tupnI 65 )% INR

( 1.5 ycneuqerf fitoM ycneuqerf 55 NC2α TFIIB 1.0 45 reP TBP 35 BRE 0.5

25 TATA 0 lortn J L 2 0PLPR 2 1 5 6KRG 2FLI 7 72S 2SK 2 9AX 2 4 9 5FA 7D 1B A AR SARR B1H1TS K C H DPE N G 9 L 1 L 4 T B2H 3 S RGE DO G YM I N D DLP 15 A2H1TSIH F CC D ECG R D L E R L C D-A P AT T 2 PR o RAC DW PC C AP D B OH DP S AG 1TS c HS >2 >1.4 >1 <1 LH 7r I I

Median fold enrichment hc H H

Fig. 6. NC2 occupancy at core promoters. (a) Relation between frequencies of core promoter elements and NC2␣ occupancy. The indicated NC2␣ enrichment classes were referenced to a survey of the core promoter composition of 9,010 annotated transcription start site sequences (24) to relate the relative frequency of TATA box (TATA), initiator (INR), downstream promoter element (DPE), TFIIB recognition element (BRE), and motif 10 element (MTE) to NC2␣ occupancy levels. (b) Core promoter regions of the indicated genes were assayed by conventional qPCR for occupancy by NC2␣ (red bars), TFIIB (green bars), and TBP (black bars). Values represent mean and standard deviation of two independent ChIP experiments. The genes are sorted from left to right to reﬂect increasing median fold enrichment values on the microarray, with the vertical line indicating the threshold of 1.2-fold enrichment (compare with SI Table 1). Chr.7 control denotes a transcriptionally inactive intergenic region on chromosome 7 that was used as a reference.

of primers, scattered throughout the gene, were used in quan- high NC2 occupancy (76%, versus the 64% in promoters with no titative real-time PCR ChIP (Fig. 5b). NC2 occupies a broad NC2␣ occupancy). We conclude that core promoter sequences region of Ϸ600 bp at the enhancer, perhaps reflecting binding of contribute to the variations in NC2␣/TFIIB ratios observed several NC2 molecules. This range is significantly above the in vivo. resolution of Ϸ200–300 bp seen at the promoter (Fig. 5b, compare signals of amplicons 2 and 3). Conversely, NC2␣ was NC2/TFIIB Ratios Vary from One Gene to Another. If core promoters not detected within the coding region. influence the binding of TFIIB and NC2, their ratios should TCR␤ was subsequently also used as a model gene to monitor differ from one gene to another. To investigate this issue and to binding of NC2␤ by a polyclonal antibody (17). Notably, the binding further confirm the ChIPchip analysis, we conducted extensive patterns of NC2␣ and NC2␤ were highly comparable to each other single-gene ChIP analysis using antibodies against NC2␣, TFIIB, (Fig. 5 b vs. c). Subsequently, the polyclonal NC2␤ antibody was also and TBP. Twenty of 21 target promoters that harbor NC2 in used to conduct two independent genome-wide promoter analyses. ChIPchip proved positive and eight of eight as true negatives, Unfortunately, under the less sensitive ChIPchip conditions, the respectively (Fig. 6b). With few exceptions, the relative levels of NC2␤ antibody proved insufficient to fully define the genes bound NC2 in single-gene and CHIPchip were well correlated. Con- by this NC2 subunit. However, by applying stringent peak analyses sistently, the relative levels of TFIIB and TBP were high at active to two independent NC2␤ ChIPchip experiments, we could define genes and low at inactive genes (SI Table 1). Furthermore, the a minimal high-confidence set of 311 target genes (SI Data Set 4), ratio between NC2 and TFIIB varied over a range of at least one which exclusively represent high-occupancy genes of NC2␣ (as an order of magnitude within this group of genes. Consistent with example for cooccupancy, the MHC-II locus is shown in SI Fig. 13). previous ChIP analysis, the NC2/TBP ratios also displayed broad

variations (14). Importantly, these data point to gene-regulatory BIOCHEMISTRY Role of Core Promoter Elements for NC2 Occupancy. Histone pro- functions of NC2 at the level of individual genes. It will be of moters usually contain TATA boxes (22), whereas the promoters interest to extend this investigation to all gene promoter regions of ribosomal protein genes are often devoid of a consensus as well as to specific cellular states. TATA element (23). Instead, ribosomal genes generally harbor a specific initiator, the so-called 5Ј-TOP (5Ј-terminal oligopyri- Discussion midine tract) element. To investigate a correlation with core We present a genome-wide study addressing the promoter sequences statistically, we referenced the ChIPchip gene set to occupancy of the general cofactor NC2 of RNAPII in eukaryotic a compendium of 9,010 annotated human promoter sequences cells. At present, the data are entirely consistent with a model in with experimentally supported transcription start sites (24). The which, upon recruitment, TBP is maintained on the gene by NC2. latter contains details about conservation and relative position- Depending on core promoter sequences and perhaps other ing of five core promoter elements, namely TATA box, initiator regulatory information encoded in cis, TFIIB, and PICs may (INR), TFIIB recognition element (BRE), downstream pro- then replace NC2–TBP complexes. The large number of genes moter element (DPE), and motif 10 element (MTE), all iden- that are bound by NC2 support the original hypothesis of a tified by a combination of positional weight matrices and con- general role of NC2 in gene transcription. Further in line with the servation between human and mouse orthologues. We could not general cofactor hypothesis, NC2 clusters roughly 50–100 bp observe any significant correlation of NC2 occupancy with the upstream of the transcription start site of at least 25% of all relative frequency of either TATA or DPE motifs (Fig. 6a). On human genes. the other hand, the BRE frequency (conserved position from Beyond it, the analysis of a large number of individual genes Ϫ37 to Ϫ32, IUPAC consensus SSRCGCC) had a reverse by single-gene ChIP analyses strongly argued against a mere correlation with NC2 occupancy. As depicted in Fig. 6a, the global function. Gene-specific contributions are suggested by the frequency of the BRE is Ϸ16% on genes with high NC2 marked differences in NC2/TFIIB ratios on different genes. occupancy compared with 35% on genes with undetectable NC2. Provided that these differences are not entirely attributable to We further observed an, albeit less pronounced, increase in INR differential epitope accessibility, the data suggest that mamma- frequency (Ϫ2toϩ4, consensus YYANWYY) at genes with lian cells differ from yeast cells that show uniform ratios of NC2␣

Albert et al. PNAS ͉ June 12, 2007 ͉ vol. 104 ͉ no. 24 ͉ 10003 Downloaded by guest on October 2, 2021 (Bur6p) with TFIIB and TBP (11). NC2 is also present on gene in vitro for binding to this and other elements proposed by regulatory enhancer regions (Fig. 5). It will be of interest to bioinformatic analysis of target genes. However, until now, we clarify whether this adds further means for gene-specific control. could not demonstrate specific binding to novel core promoter NC2 occupancy positively correlated with mRNA levels and elements other than the TATA box, which itself is recognized negatively with core promoter binding sites for TFIIB (BRE). with limited specificity (14). Whereas the latter is nicely explained by the previously noted The two NC2 subunits tightly interact through histone folds competitive binding of NC2 and TFIIB to TBP–promoter com- both in vitro and in vivo (ref. 17 and M.M., unpublished obser- plexes (5, 6), the observed positive correlation with steady-state vations). Nonetheless, it was proposed that the two individual mRNA levels is counterintuitive for a repressor. It is, however, subunits could have separate functions, even exist in free forms in line with the ChIP and the genetic analyses in yeast (8, 11, 13, or in other yet-to-be-defined complexes in yeast (36). At present, 25) and, less developed, also in human cells (26). Dominant our ChIP data show good correlation of both subunits at selected negative versions of NC2 were inhibitory to the growth of individual genes (e.g., TCR␤). Moreover, the ChIPchip data transformed cells (E. Piaia and M.M., unpublished observa- define a limited set of genes harboring both NC2 subunits. tions), preventing a functional classification of target genes. Further studies (i.e., Re-ChIP, tagged NC2 proteins) will be However, direct functional interpretations of gene occupancy required to more systematically address the question of which data are often questionable. Factors may be acting in the present genes harbor the NC2 complex or individual subunits. situation or, in fact, preparing the gene for a later situation. In Finally, we noted that NC2/TFIIB ratios remained constant general, the complex control of transcription at multiple levels during cell cycle-dependent activation of histone genes. Similar complicates the interpretation. For example, a gene might be observations were made in the context of more sensitive acti- loaded with positively acting general transcription factors, al- vation models such as the IL-2 gene in T cells (T.K.A. and M.M., though efficient elongation of RNAPII is inhibited by negative unpublished data). Here, TBP levels also rose together with NC2 elongation factors (27). Hence, genome-wide occupancy data and TFIIB during activation. It will be of interest to clarify occasionally contradict gene-regulatory functions. Examples are whether TBP-NC2 and TBP/TFIIB (which is equivalent to the T-bet (and other gene regulatory factors), which occupies pro- PIC) represent two main alternative states that genes adopt at moters that are not under its control (28), RNAPII, which poises the level of preinitiation complexes in mammalian cells. genes for activation by binding them in an inactive state (29), and Materials and Methods Mediator complexes, which locate to coding regions where they lack a known function (30). Last but not least, steady-state Cell Culture. EBV-transformed lymphoblastoid B-cell line mRNA levels depend on transcription as well as downstream LCL721, T cell line Jurkat (subclone J6), and Burkitt’s lymphoma cell line Raji were maintained in RPMI medium 1640 events such as RNA processing, RNA export, and RNA stability. growth medium supplemented with 10% FBS, 2 mM L- Nevertheless, the ChIP results may be interpreted within the glutamine and 100 units/ml penicillin-streptomycin (all by In- defined molecular properties of the NC2 complex. For example, vitrogen, Carlsbad, CA) at 37°C with 5% CO . Cells were the correlation of steady-state mRNA levels with NC2 occu- 2 harvested at a density of 0.5 to 1 ϫ 106 cells per milliliter. pancy may relate to a marked stabilization of TBP complexes on Centrifugal elutriation (Beckman, Fullerton, CA) was used to TATA-containing genes (M.M., unpublished observation). On enrich Raji cell populations in different cell cycle phases. the other hand, in vitro transcription analysis of selected highly Individual fractions were counted and their DNA content de- active genes (MYC, histone H2B, and RPS27), displaying high termined by flow cytometry (Becton Dickinson, Franklin Lakes, NC2 occupancy in B cells, suggested that their transcription is NJ) by using standard procedures. inhibited by NC2. Consistent with earlier findings (7), a viral promoter is negatively controlled by NC2 in both the presence ChIP and ChIPchip Analysis. ChIP and ChIPchip assays were per- and the absence of the activator VP16 (SI Fig. 14). These data formed essentially as described (37). Per single ChIP, 100 ␮lof further underline our previous conclusions that NC2 acts as a chromatin extract (1 mg/ml DNA) were incubated with 2–5 ␮g repressor in vitro. However, given that genes have to maintain of antibodies. DNA from ChIP and input samples was subse- their epigenetic status for long periods of time in cells, the quently purified and resuspended in 100 ␮l of 10 mM Tris⅐HCl maintenance of TBP mediated by NC2 may perhaps lead to (pH 8.0). For real-time quantitative PCR (qPCR) analysis, 2 ␮l positive effects on gene expression in living cells. Such effects of ChIP and input DNA were used as template. For ChIPchip may not be revealed in the in vitro transcription analyses, which analyses, up to 200 ng of ChIP and input DNA were end- monitor predominantly the initiation process and the first round polished, ligated to linkers, and amplified. of transcription. Based on these in vitro data, we assume that genes, although qPCR Analysis. qPCR was performed by using a SYBR green potentially being poised for activation, remain in a repressed detection system on an ABI 5700 thermal cycler (Applied state until NC2 is removed. Both biochemical and x-ray data Biosystems, Foster City, CA). Reactions were carried out in suggested that NC2 and TFIIB compete for binding to TBP– 25-␮l volume containing 10 pmol of each primer and 2 ␮lof DNA complexes (5, 6, 31). The antagonism with TFIIA, al- serially diluted input or ChIP DNA as templates. Primer se- though controversially discussed at the level of the x-ray struc- quences are available upon request. ture (31, 32), is supported by both biochemical and genetic data (5, 7). Other cofactors such as BTAF1/Mot1 may contribute to Total RNA Preparation and RT-PCR Analysis. Total RNA was pre- the removal of NC2 (26, 33). The equilibrium with TFIIB pared with the RNeasy Midi kit (Qiagen, Valencia, CA) accord- provides a reasonable explanation for the relative absence of ing to the manufacturer’s recommendation. Reverse transcrip- BRE elements on NC2 genes with high occupancy. Histone tion used Thermoscript enzyme (Invitrogen) with 2 ␮g of total genes that usually have TATA but lack BREs as well as RNA, random hexameric primers for first-strand synthesis, and ribosomal genes that generally lack BREs and occasionally gene-specific primer pairs for PCR amplification of cDNAs. harbor TATA in their promoters fall in this category. Of further note, the controversial binding of NC2 to DPE elements (34, 35) DNA Microarrays. For ChIPchip, promoter tiling arrays were used was not confirmed by our genome-wide promoter analysis. that cover 24,275 human promoter regions (NimbleGen Systems, Instead, NC2 occupancy correlates slightly with the presence of Madison, WI). For each region, 15 50-mer probes are tiled with an intact initiator. We tested recombinant TBP–NC2 complexes Ϸ100-bp spacing over a 1.5-kb sequence window around TSSs.

10004 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0703490104 Albert et al. Downloaded by guest on October 2, 2021 For gene expression profiling, high-density oligonucleotide ar- We thank the members our laboratory for discussions. This work was rays for the analysis of Ͼ47,000 human transcripts were used supported by Federal Ministry of Education and Research (BMBF) (Human Genome U133 Plus 2.0 Array, Affymetrix, Santa Clara, Grant 0313030D (to K.G.) and by Deutsche Forschungsgemeinschaft CA). Sample labeling, hybridization, scanning, raw data extrac- Grant SFB646, the Specific Targeted Research Project (STREP) pro- tion, and analyses were conducted by an authorized Affymetrix gram of the European Union Grant HPRN-CT-2002-00261, and BMBF service provider (KBF, Regensburg, Germany). Grants 0313030A and 0313427 (to M.M.).

1. Roeder RG (1996) Trends Biochem Sci 21:327–335. 19. Eisenberg E, Levanon EY (2003) Trends Genet 19:362–365. 2. Thomas MC, Chiang CM (2006) Crit Rev Biochem Mol Biol 41:105–178. 20. Heintz N, Roeder RG (1984) Proc Natl Acad Sci USA 81:2713–2717. 3. Meisterernst M, Roeder RG (1991) Cell 67:557–567. 21. Christova R, Oelgeschlager T (2002) Nat Cell Biol 4:79–82. 4. Inostroza JA, Mermelstein FH, Ha I, Lane WS, Reinberg D (1992) Cell 22. Stein GS, Stein JL, van Wijnen AJ, Lian JB (1992) Curr Opin Cell Biol 70:477–489. 4:166–173. 5. Goppelt A, Stelzer G, Lottspeich F, Meisterernst M (1996) EMBO J 15:3105– 23. Perry RP (2005) BMC Evol Biol 5:15. 3116. 24. Jin VX, Singer GA, Agosto-Perez FJ, Liyanarachchi S, Davuluri RV (2006) 6. Mermelstein F, Yeung K, Cao J, Inostroza JA, Erdjument-Bromage H, BMC Bioinformatics 7:114. Eagelson K, Landsman D, Levitt P, Tempst P, Reinberg D (1996) Genes Dev 25. Lemaire M, Xie J, Meisterernst M, Collart MA (2000) Mol Microbiol 36:163– 10:1033–1048. 173. 7. Xie J, Collart M, Lemaire M, Stelzer G, Meisterernst M (2000) EMBO J 26. Klejman MP, Pereira LA, van Zeeburg HJ, Gilfillan S, Meisterernst M, 19:672–682. Timmers HT (2004) Mol Cell Biol 24:10072–10082. 8. Prelich G (1997) Mol Cell Biol 17:2057–2065. 27. Sims RJ, III, Mandal SS, Reinberg D (2004) Curr Opin Cell Biol 16:263–271. 9. Gadbois EL, Chao DM, Reese JC, Green MR, Young RA (1997) Proc Natl 28. Beima KM, Miazgowicz MM, Lewis MD, Yan PS, Huang TH, Weinmann AS Acad Sci USA 94:3145–3150. (2006) J Biol Chem 281:11992–12000. 10. Lee TI, Wyrick JJ, Koh SS, Jennings EG, Gadbois EL, Young RA (1998) Mol 29. Radonjic M, Andrau JC, Lijnzaad P, Kemmeren P, Kockelkorn TT, van Cell Biol 18:4455–4462. Leenen D, van Berkum NL, Holstege FC (2005) Mol Cell 18:171–183. 11. Geisberg JV, Holstege FC, Young RA, Struhl K (2001) Mol Cell Biol 30. Andrau JC, van de Pasch L, Lijnzaad P, Bijma T, Koerkamp MG, van de Peppel 21:2736–2742. J, Werner M, Holstege FC (2006) Mol Cell 22:179–192. 12. Cang Y, Prelich G (2002) Proc Natl Acad Sci USA 99:12727–12732. 31. Kamada K, Shu F, Chen H, Malik S, Stelzer G, Roeder RG, Meisterernst M, 13. Chitikila C, Huisinga KL, Irvin JD, Basehoar AD, Pugh BF (2002) Mol Cell Burley SK (2001) Cell 106:71–81. 10:871–882. 32. Bleichenbacher M, Tan S, Richmond TJ (2003) J Mol Biol 332:783–793. 14. Gilfillan S, Stelzer G, Piaia E, Hofmann MG, Meisterernst M (2005) J Biol 33. Geisberg JV, Moqtaderi Z, Kuras L, Struhl K (2002) Mol Cell Biol 22:8122– Chem 280:6222–6230. 8134. 15. Ren B, Robert F, Wyrick JJ, Aparicio O, Jennings EG, Simon I, Zeitlinger J, 34. Willy PJ, Kobayashi R, Kadonaga JT (2000) Science 290:982–985. Schreiber J, Hannett N, Kanin E, et al. (2000) Science 290:2306–2309. 35. Lewis BA, Sims RJ, III, Lane WS, Reinberg D (2005) Mol Cell 18:471–481. 16. Kim TH, Barrera LO, Zheng M, Qu C, Singer MA, Richmond TA, Wu Y, 36. Creton S, Svejstrup JQ, Collart MA (2002) Genes Dev 16:3265–3276. Green RD, Ren B (2005) Nature 436:876–880. 37. Odom DT, Zizlsperger N, Gordon DB, Bell GW, Rinaldi NJ, Murray HL, 17. Gilfillan S, Stelzer G, Kremmer E, Meisterernst M (2003) Methods Enzymol Volkert TL, Schreiber J, Rolfe PA, Gifford DK, et al. (2004) Science 303:1378– 370:467–479. 1381. 18. Bieda M, Xu X, Singer MA, Green R, Farnham PJ (2006) Genome Res 38. Nikolov DB, Chen H, Halay ED, Usheva AA, Hisatake K, Lee DK, Roeder 16:595–605. RG, Burley SK (1995) Nature 377:119–128. BIOCHEMISTRY

Albert et al. PNAS ͉ June 12, 2007 ͉ vol. 104 ͉ no. 24 ͉ 10005 Downloaded by guest on October 2, 2021