Lineage-Specific Programming Target Genes Defines Potential for Th1 Temporal Induction Pattern of STAT4

Temporal Induction Pattern of STAT4 Target Genes Defines Potential for Th1 Lineage-Specific Programming

This information is current as Seth R. Good, Vivian T. Thieu, Anubhav N. Mathur, Qing of September 26, 2021. Yu, Gretta L. Stritesky, Norman Yeh, John T. O'Malley, Narayanan B. Perumal and Mark H. Kaplan J Immunol 2009; 183:3839-3847; Prepublished online 26 August 2009;

doi: 10.4049/jimmunol.0901411 Downloaded from http://www.jimmunol.org/content/183/6/3839

Supplementary http://www.jimmunol.org/content/suppl/2009/08/26/jimmunol.090141 Material 1.DC1 http://www.jimmunol.org/ References This article cites 35 articles, 19 of which you can access for free at: http://www.jimmunol.org/content/183/6/3839.full#ref-list-1

Why The JI? Submit online.

• Rapid Reviews! 30 days* from submission to initial decision by guest on September 26, 2021

• No Triage! Every submission reviewed by practicing scientists

• Fast Publication! 4 weeks from acceptance to publication

*average

Subscription Information about subscribing to The Journal of Immunology is online at: http://jimmunol.org/subscription Permissions Submit copyright permission requests at: http://www.aai.org/About/Publications/JI/copyright.html Email Alerts Receive free email-alerts when new articles cite this article. Sign up at: http://jimmunol.org/alerts

The Journal of Immunology is published twice each month by The American Association of Immunologists, Inc., 1451 Rockville Pike, Suite 650, Rockville, MD 20852 Copyright © 2009 by The American Association of Immunologists, Inc. All rights reserved. Print ISSN: 0022-1767 Online ISSN: 1550-6606. The Journal of Immunology

Temporal Induction Pattern of STAT4 Target Genes Deﬁnes Potential for Th1 Lineage-Speciﬁc Programming1

Seth R. Good,2* Vivian T. Thieu,2† Anubhav N. Mathur,† Qing Yu,† Gretta L. Stritesky,† Norman Yeh,† John T. O’Malley,† Narayanan B. Perumal,3* and Mark H. Kaplan3†

STAT4 is a critical component in the development of inflammatory adaptive immune responses. It has been extensively characterized as a lineage-determining factor in Th1 development. However, the genetic program activated by STAT4 that results in an inflammatory cell type is not well defined. In this report, we use DNA isolated from STAT4-chromatin immunoprecipitation to perform chromatin immunoprecipitation-on-chip analysis of over 28,000 mouse gene promoters to identify STAT4 targets. We demonstrate that STAT4 binds multiple gene-sets that program distinct components of the Th1 lineage. Although many STAT4 target genes display STAT4-dependent IL-12-inducible expression, other genes displayed IL-12-induced histone modifications but lack induction, possibly due to high relative basal expression. In the subset of genes that STAT4 programs for expression in Th1 Downloaded from cells, IL-12-induced mRNA levels remain increased for a longer time than mRNA from genes that are not programmed. This suggests that STAT4 binding to target genes, while critical, is not the only determinant for STAT4-dependent gene programming during Th1 differentiation. The Journal of Immunology, 2009, 183: 3839–3847.

he Th1 cell is responsible for cell-mediated immune tin immunoprecipitation(ChIP)4-on-chip) or large-scale se-

functions including eradication of intracellular patho- quencing of the bound targets (ChIP-seq) to identify, charac- http://www.jimmunol.org/ gens. STAT4 is required for the development of Th1 terize, and analyze transcription of genes controlled by T ϩ cells from naive CD4 T cells and most IL-12-stimulated func- transcription factors (13, 14). These studies provide a global tions (1, 2). STAT4-deficient mice are highly susceptible to picture of transcription factor-target gene interactions in the infection by intracellular pathogens but are resistant to T cell- physiological state. Such a ChIP-on-chip study may reveal mediated autoimmune diseases (1). In humans, STAT4 also me- STAT4-mediated transcriptional regulatory networks active in diates IL-12 functions and single nucleotide polymorphisms in Th1 cell development. the STAT4 gene correlate with susceptibility to autoimmune In this report, we have performed ChIP-on-chip experiments for disease (3, 4). How STAT4 establishes the Th1 genetic program STAT4. Using informatics and biological interrogation of the data, remains unclear. we have identified additional aspects of the STAT4-dependent ge- by guest on September 26, 2021 Following activation, STAT4 binds to cis-regulatory regions netic program and activation patterns that predict potential pro- of several genes to induce gene expression, though relatively gramming of Th1-specific gene expression. Moreover, we identify few of these have been directly identified. In humans STAT4 additional STAT4 target genes that may play a role in the inflam- has been shown to bind to IFNG and IL12RB2, and in mouse matory cell phenotype. These data establish a framework for ge- Ifng, Hlx1, and Il18r1 (5–10). However, a much larger set of nome-wide understanding of the transcription factor network that genes appears to rely on STAT4 for expression in the Th1 dif- regulates inflammatory immune responses. ferentiation program (9, 11, 12). It is possible that the ability of STAT4 to activate an inflammatory genetic program can be Materials and Methods understood better by a genome-wide analysis of target gene T cell preparation and analysis promoters. Recently, a number of studies have used chromatin For ChIP-on-chip experiments CD4ϩ T cells from C57BL/6 mice were acti- immunoprecipitation followed by microarray analysis (chroma- vated for 3 days with anti-CD3 and anti-CD28. Cells were washed and incubated in the presence or absence of 5 ng/ml IL-12 for 4 h before being pro- cessed for ChIP analysis by Genpathway. For analysis of gene induction and histone modification, wild-type and Stat4Ϫ/Ϫ BALB/c mice were activated for *School of Informatics, Indiana University-Purdue University Indianapolis; and 3 days as above before RNA was isolated from cells that were unstimulated or †Departments of Pediatrics, H.B. Wells Center for Pediatric Research and De- stimulated with 5 ng/ml IL-12 for 4 or 18 h. Generation of Th1 cultures and partment of Microbiology and Immunology, Indiana University School of Med- quantitative (real-time) PCR (qPCR) was performed as described (15). ELISA icine, Indianapolis, IN 46202 for TNF-␣ was performed using reagents from BD Pharmingen. DNA affinity Received for publication May 5, 2009. Accepted for publication July 16, 2009. precipitation assay and ChIP was performed as described (8). ERK immuno- blots were performed using standard methods (8). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance FactorPath ChIP-on-chip with 18 U.S.C. Section 1734 solely to indicate this fact. 1 This work was supported by Public Health Service Grant AI45515 (to M.H.K.). G.L.S., Preparation of cells for ChIP-on-chip was performed at GenPathway (16). N.Y., and J.T.O. were supported by T32AI060519; V.T.T. was supported by T32HL007910. Cells prepared as above were fixed with 1% formaldehyde for 15 min and quenched with 0.125 M glycine. Cell lysates were sonicated to an average 2 S.R.G. and V.T.T. contributed equally to this study. 3 Address correspondence and reprint requests to Dr. Mark H. Kaplan, Depart- ments of Pediatrics, and Microbiology, and Immunology, Indiana University 4 Abbreviations used in this paper: ChIP, chromatin immunoprecipitation; qPCR, School of Medicine, H.B. Wells Center for Pediatric Research, 702 Barnhill quantitative (real-time) PCR; GO, gene ontology; TSS, transcription start site; Drive, RI 2600, Indianapolis, IN 46202 or Dr. Narayanan B. Perumal, School of MEME, multiple expectation maximization for motif elicitation. Informatics, Indiana University-Purdue University Indianapolis, Indianapolis, IN 46202. E-mail address: [email protected] or [email protected] Copyright © 2009 by The American Association of Immunologists, Inc. 0022-1767/09/$2.00 www.jimmunol.org/cgi/doi/10.4049/jimmunol.0901411 3840 STAT4 TARGET GENES IN Th1 CELLS Downloaded from http://www.jimmunol.org/ FIGURE 1. STAT4 binding in target gene promoters. A, Wild-type CD4ϩ T cells were activated for 3 days and cultured in the presence or absence of IL-12 for 4 h. ChIP analysis was performed for STAT4 or control IgG and qPCR was performed to determine STAT4 binding at the indicated target sequences. The scale indicates relative binding that is normalized to input DNA. B, The number of genes was graphed against binding intensity. Intensity of 4669 intervals ranged from 2.2–32.4. The cut-off intensity for subsequent analysis was 4. C, Histograms of the frequency of STAT4 binding across the arrayed promoter sequences relative to the TSS. The curated list has all 1540 genes with binding intensity Ͼ4. The exact match list is a subset of 660 genes that contain a consensus STAT4 binding site in the promoter. D, Comparison of the frequency of subsets of genes in the curated list (binding intensity Ͼ4) and the frequency on the Affymetrix chip using GOstat to assign gene function. p values for significant differences are indicated. E, Venn diagram comparing the overlap of IL-12-induced genes from published microarray analysis (see text) with the ChIP-on-chip dataset. by guest on September 26, 2021

DNA fragment length of 300–500 bp. Genomic DNA from aliquots was under Th1 conditions for 0, 2, 6, and 48 h (MG-U74Av2 arrays containing purified for use as input. Chromatin was precleared with protein A agarose 12,488 probes) (12), differentiated C57BL/6 mouse Th1 cells stimulated beads (Invitrogen) and STAT4-bound DNA sequences were isolated using with IL-12 for 18 h (custom array with 24,000 probes) (11), differentiated specific Abs (Santa Cruz sc-486). Complexes were precipitated with pro- human Th1 cells stimulated with IL-12 for 4 or 24 h (HuGeneFL array tein A agarose beads, washed, eluted from the beads with SDS buffer, and containing 6000 genes) (17), and PHA/IL-2 activated human T cells stim- subjected to RNase and proteinase K treatment. Crosslinks were reversed ulated with IL-12 for 6 h (U133A and U133B arrays) (18). by incubation overnight at 65°C, and ChIP DNA was purified by phenol- chloroform extraction and ethanol precipitation. qPCR reactions were con- Multiple expectation maximization for motif elicitation (MEME) ducted in triplicate on specific genomic regions using SYBR Green Su- analysis permix (Bio-Rad). The resulting signals were normalized for primer Ͼ efficiency by carrying out quantitative PCR for each primer pair using The 1111 interval sequences with peak intensity 4, which were in close input DNA. proximity to 1540 genes, were screened by MEME (19) to find over-rep- Analysis of ChIP DNA by Affymetrix tiling arrays was conducted by resented motifs. The parameters for MEME included looking at the reverse GenPathway. ChIP (multiple reactions) and input DNAs were amplified complement sequence, a maximum motif width of 15, and a cutoff of 30 using ligation-mediated PCR. The resulting amplified DNA was purified, motifs ordered by increasing E-value. The analysis was run using three quantified, and tested by qPCR at the same specific genomic regions as the different approaches regarding the number of times the motif repeated, any original ChIP DNA to assess quality of the amplification reactions. Am- number of repeats, zero or one repeat, and one or more repeats with vir- plified DNA was fragmented and labeled using the DNA Terminal Label- tually the same results each time. WebLogo (20) generated the motif logos ing Kit from Affymetrix, and hybridized to the Affymetrix GeneChip using the aligned site occurrences for each motif given by MEME. Mouse Promoter 1.0R Array at 45°C overnight. This array contains over Gene ontology (GO) analysis 4.6 million 25-mer oligos tiled to interrogate over 28,000 mouse promoter regions (Ϫ7.5 kb to ϩ2.5kb relative to the transcription start site). Arrays The GO analysis was performed by GOstat (21). Using an input list and a were scanned and the resulting CEL files were analyzed using Affymetrix selected background, GOstat determines enriched GO terms based on a TAS software that computed estimates of fold enrichment over hybrid- Fisher exact test. The p-value from the Fisher exact test is altered using a ization with input DNA in linear scale. This metric is referred to as peak Benjamini correction. Only terms with a corrected p-value less than 0.1 are binding intensity. Thresholds were selected, and the resulting BED files reported. The analysis was performed three times with the curated gene list containing lists of sequence intervals with significant binding detected as the input and the Affymetrix chip gene list as the background. were analyzed using Genpathway software that provides comprehensive information on genomic annotation, peak metrics, and sample compar- Results isons for all peaks (intervals). Graphing of peak binding values were generated using the Affymetrix Integrated Genome Browser. The entire dataset Generation and analysis of STAT4 ChIP-on-chip data is submitted to the NCBI GEO database (GSE16845; www.ncbi.nlm. To identify STAT4 target genes, CD4ϩ T cells were activated for nih.gov/geo/query/acc.cgi?accϭGSE16845). Intersects of our dataset with previously published datasets were deter- three days with anti-CD3 before washed cells were stimulated with mined using Microsoft Excel. Published datasets that were used for anal- IL-12 for 4 h and chromatin was immunoprecipitated with anti- ysis of common genes were from naive BALB/c mouse Th cells cultured STAT4 Ab. The preactivation was performed to allow cells to The Journal of Immunology 3841 acquire IL-12 responsiveness (22). IL-12 stimulation of activated cells resulted in 70–80% of cells in the population becoming pSTAT4-positive with a 6-fold increase in mean fluorescence intensity, as assessed by intracellular staining (data not shown). To confirm STAT4 binding to target genes, we tested Il18r1 promoter and the Il2ra PRRIII element DNA in the precipitate following IL-12 stimulation (8, 23) and observed a greater than 30-fold induction of STAT4 binding, compared with unstimulated cells (Fig. 1A). DNA from the ChIP was then hybridized to an Affymetrix array containing sequences spanning Ϫ7.5kB to ϩ2.5kb of over 28,000 promoters in the mouse genome and enrichment was calculated based on comparison to an array hybridized with input DNA. In replicate experiments, the ChIP-on-chip analysis identified a total of 4,669 genes as near a STAT4 binding site in vivo. The peak binding intensities (see definition in Materials and Meth- ods) of immunoprecipitated DNA hybridizing to the promoter oligonucleotides showed a wide range (2.2 to 32.4, with the highest

value being the most intense binding) indicating the diverse nature Downloaded from of STAT4 binding to target genes. Mean and median of this data set were at 4.4 and 3.3, respectively (Fig. 1B). A manual inspection of the data showed many of the known STAT4 target and Th1- associated genes to have peak intensities of four or higher. Thus, we ﬁltered the list of genes using a cutoff peak intensity of 4 (1111

binding sites corresponding to 1540 genes) referred to as the http://www.jimmunol.org/ “curated” list and all subsequent analyses were performed on this curated gene set (Table S1).5 We confirmed binding of STAT4 to genes with a range of peak intensities from 4.5 to 21 using ChIP assay in wild-type and Stat4Ϫ/Ϫ activated T cells. The enrichment of percent input between wild-type and Stat4Ϫ/Ϫ samples was similar to the enrichment observed between IL-12 stimulated and unstimulated cells (Figs. 1A and 2). These results further confirm the identification of STAT4 targets in this dataset. We then analyzed the frequency of STAT4 binding between 7.5 by guest on September 26, 2021 kb upstream to 2.5 kb downstream of the transcription start site (TSS) to define chromatin location preferences for STAT4. A pre- ponderance of STAT4 binding is in and around the target gene TSS (Fig. 1C). There was no difference in frequency between the curated gene list and a subset of genes that had an exact STAT4 consensus motif (TTCNNNGAA in 660 of 1111 binding sites or 59.4%) (Fig. 3A) in the association of STAT4 with specific regions in the promoter. To determine whether STAT4 target genes were enriched for any specific category, we defined the enrichment of GO terms in the curated set compared with an Affymetrix mouse genome background using the GOstat program (21). The major statistically sig- Ϫ/Ϫ nificant enrichments ( p Ͻ 0.05) were in genes involved in cyto- FIGURE 2. STAT4 binding to target genes. WT and Stat4 activated kine binding, including factors with IL receptor activity and TNF T cells were stimulated with IL-12 for 4 h before nuclei were isolated for chromatin immunoprecipitation with anti-STAT4. Results are shown as receptor binding activity (Fig. 1D). There was also a significant percent input as calculated using qPCR results of amplification with prim- enrichment for nucleic acid binding factors. These analyses dem- ers specific for the promoter region of each of the genes indicated. The onstrate selectivity in STAT4 binding target genes associated with peak intensity of STAT4 binding derived from the ChIP-on-chip dataset specific cellular functions. and the fold-enrichment of STAT4 binding in WT, compared with Ϫ Ϫ Ϫ Ϫ A number of published reports have examined IL-12-induced Stat4 / cells, are indicated. STAT4 binding in the Stat4 / cells was gene expression using microarrays from human or mouse cells within 1 SD of the background (IgG control). treated with IL-12 for various times during or after Th1 differentiation (11, 12, 17, 18). We generated a cumulative list of IL-12- induced genes from these studies and compared them to the ChIP- Computational identification of the STAT4 consensus motif on-chip dataset. The Venn diagram in Fig. 1E identifies an overlap To define the proportion of STAT4-bound genes that contained of 72 genes (Table S2A) between these two analyses. a STAT4 consensus motif, TTCNNNGAA (24), we analyzed the curated gene set by directly examining consensus sites using a motif search algorithm. A Perl script to identify exact matches allowing the ‘NNN’ degeneracy at positions 4–6 identified mo- 5 The online version of this article contains supplemental material. tif 1 shown in Fig. 3A at a frequency of 59.4% in the curated 3842 STAT4 TARGET GENES IN Th1 CELLS Downloaded from http://www.jimmunol.org/

FIGURE 3. Motif analysis of STAT4 target genes. A, WebLogo representations of STAT4 consensus sequences identified by direct search for the TTCNNNGAA consensus (motif 1) or sequences identified by MEME motif search (motifs 2–5). The percentage of genes/intervals that contain sequences related to each motif is indicated with E-values for MEME motifs. B, Percentage of sequences in each range of peak values that contain a consensus STAT binding site (A, motif 1). Note that as percentages of sequences with the consensus decreases, the total number of sequences in a range increases (compare with Fig. 1B). C, Histogram of the proximity of motifs 4 and 5 to the consensus STAT motif 2 graphed as the frequency of each motif occurring in 100 bp intervals from the STAT consensus. D, DNA affinity purification analysis (DAPA) of motif 3 using biotinylated oligonucleotides. Total cell extracts isolated from purified CD4 T cells activated for 3 days and left unstimulated or stimulated for 4 h with IL-12 or IL-18 as indicated were incubated with by guest on September 26, 2021 oligonucleotides before complexes were precipitated with streptavidin agarose. Precipitated proteins were visualized using immunoblot. E, DAPA analysis as in C with the addition of competition (fold molar excess indicated) of binding with a STAT consensus (motif 3) or a nonspecific competitor. gene list. In the second approach, we used the expectation- motif 5 occurrences within 100 bp intervals relative to the po- maximization motif search algorithm, MEME (19) to identify sition of the consensus STAT site (motif 2). Motifs 4 and 5 were motifs. We observed a number of motifs with low E-values. enriched in proximity to the STAT site, almost entirely within Sequences containing variants of a STAT4 consensus site were 1 kb of the consensus sequence (Fig. 3C). We functionally as- found in all STAT4 bound sequences (Fig. 3A, motif 2), further sessed binding to motif 3 using biotinylated oligonucleotides validating the dataset and suggesting that the false-positive rate that spanned the motif. We detected STAT4 binding specifically is low. The percentage of the genes containing the consensus to motif 3 that was competed by STAT consensus oligonucle- (motif 1) was highest in sequences that had high peak values, otides to a greater extent than by nonspecific sequences (Fig. 3, and decreased as peak values decreased (Fig. 3B). Although D and E). these data shows that some sequences containing the STAT consensus site were not included in the curated dataset (peak STAT4 binds multiple genes associated with the Th1 phenotype value less than 4), the problem of the exclusion of these false- Although the Th1 genetic program is known to be largely de- negatives is offset by avoiding the inclusion of large numbers of pendent on STAT4, and STAT4 binding to a subset of Th1 genes (Fig. 1B) that would lack other identifiable motifs. If all genes has been demonstrated, all targets of STAT4 required to bound genes in the array are considered, the percentage with the establish Th1 differentiation are not clear. We examined the consensus falls to 43%, and the MEME algorithm identified curated list for genes demonstrated in the literature to be func- Ͻ10% of the genes as having motif 2 if the cut-off was reduced tionally associated with the Th1 phenotype, or preferentially to a peak value of 3. Thus, our analysis minimizes false-posi- expressed in Th1 cells, and have demonstrated STAT4- or T- tives at the expense of potentially excluding a limited number bet-dependence in their expression (Table S3). We identified of valid targets. binding to a set of genes (18/28 Th1 genes searched (Fig. 4A, Three additional motifs shown in Fig. 3A were obtained from Table S3); significantly enriched from the genome p Ͻ 10Ϫ8) the MEME analysis that were present in 5–10% of the total including Furin, Ifng, and Il18r1, genes that are known to be sequences and had low E-values. Manual and JASPAR-medi- STAT4 targets (5, 6, 8, 26) as well as Th1 genes that are ex- ated (25) sequence analysis identified a STAT-like site in motif pressed independently of STAT4 including Egr2 and Ms4a4b. Using 3, an NF-␬B site in motif 4 and a PPAR␥/RXR site in motif 5. Affymetrix’s Integrated Genome Browser we mapped the peak inten- These motifs were not significantly enriched in any GO cate- sities of binding in a subset of these genes (Fig. 4B). Although the gorized gene set. We mapped the frequency of motif 4 and microarray in these experiments is focused on promoter sequences, The Journal of Immunology 3843

FIGURE 4. STAT4 binding in Th1 genes. A, Heatmap of Th1 associated genes based on binding intensity of STAT4 to gene promoters. B, The In- tegrated Genome Browser (IGB) used

BAR files generated from the ChIP-on- Downloaded from chip data to identify binding peaks relative to annotated genes. The graphs indicate the significance (p-value prob- ability) of enrichment at promoter-arrayed sites. C, Wild type and Stat4Ϫ/Ϫ CD4ϩ T cells were cultured under Th1 conditions for five days before RNA http://www.jimmunol.org/ was isolated for qPCR using the indicated primers. RNA levels are relative to wild type Th1 cells. by guest on September 26, 2021

STAT4 peak binding was largely localized to the gene regions im- served a set of genes that are associated with TCR signaling iden- mediately upstream of transcription initiation. The exceptions to this tified from the literature (29) (7/23 TCR signaling genes searched included Gadd45g, which had peaks both 5-prime and 3-prime of the (Table S2B); significant enrichment from the genome p ϭ 1 ϫ gene, and Tnf that had intragenic peak binding, closely linked to 10Ϫ4) including Cd3z, Prkcq/PKC␪, and Lcp2/SLP-76 (Fig. 5A) STAT4 binding at the Lta promoter (Fig. 4B). that are important for the stimulation of T cells in response to Several of these genes, while associated with Th1 function, specific Ag. To test whether expression of these genes was affected had not been demonstrated to be STAT4-dependent. To test a by STAT4-deficiency, we differentiated wild-type and Stat4Ϫ/Ϫ requirement for STAT4 in Th1 expression of these genes, we Th1 cells and used qPCR to examine relative levels of gene ex- Ϫ/Ϫ ϩ differentiated wild-type and Stat4 CD4 T cells under Th1 pression, using Hlx1 as a STAT4-dependent control. Although conditions and used quantitative PCR to examine relative ex- expression of Prkcq and a number of other TCR signaling genes pression. Gadd45g, which plays a role in IFN-␥ production and Ϫ Ϫ was not significantly decreased in Stat4 / Th1 cells, Lcp2 ex- potentially synergy between IL-12 and IL-18 (27), was de- Ϫ/Ϫ Ϫ Ϫ pression in Stat4 cultures was decreased to one-half the creased more than 5-fold in Stat4 / Th1 cultures compared level seen in wild type cultures (Fig. 5B). To determine whether with wild-type cultures (Fig. 4C). Myd88, which we have previously shown to be IL-12-inducible in a STAT4-dependent there are consequences of this decrease in expression, we ana- manner (28), also required STAT4 for maximal expression in lyzed the TCR-induced phosphorylation of the ERK MAPKs Ϫ/Ϫ Th1 cells (Fig. 4C). and observed decreased ERK phosphorylation in Stat4 Th1 cells, compared with wild-type cells (Fig. 5C). These observa- STAT4 binds a subset of genes associated with TCR signaling tions agree with previous data suggesting that strong stimuli Ϫ Ϫ In a manual examination of the curated gene list, we defined sub- minimize the defect in IFN-␥ production by Stat4 / Th1 cells sets of genes that would have specific functional consequences for (30), though it is difficult to test using IFN-␥ as a readout be- T cell differentiation or function. Among these clusters we ob- cause STAT4 plays an important role in programming Ifng for 3844 STAT4 TARGET GENES IN Th1 CELLS

manually based on involvement in Th differentiation, immune responses or gene expression, in wild-type and Stat4Ϫ/Ϫ activated T cells stimulated for four or 18 h with IL-12. In Fig. 6A, an analysis of genes that demonstrated STAT4-dependent IL-12 induction indicated that there is no correlation between binding intensity and level of gene induction. This is exempliﬁed by Il18rap, which was one of the highest intensity STAT4 binding genes, but was induced only 2-fold at 4 h, while Il18r1, which bound with less intensity, was induced 5-fold. The three genes that showed over 20-fold induction, Furin, Ifng, and Il24, had intensity ranging from 7.5 to 21. We also performed this analysis using an IL-12-induced microarray dataset (18), and observed a similar pattern when the published fold-induction was graphed against our binding intensity values (Fig. 6B). To further characterize this subset of genes, we divided genes into categories based on the patterns of gene induction by IL-12. “Noninduced” genes had Յ2-fold induction at four or 18 h; “Tran- sient” genes had Ͼ2-fold induction at4hbutϽ2-fold induction at

18 h, and “sustained” genes had Ͼ2-fold induction at 4 and 18 h. Downloaded from For each subset we graphed the fold IL-12-induced gene expression vs the basal gene expression normalized to the endogenous ␤ control 2-microglobulin. We observed a general trend that genes with lower basal expression had greater induction, and observed that those STAT4 target genes not induced by IL-12 were among

the most highly expressed genes that we analyzed (Fig. 6C). This http://www.jimmunol.org/ suggested that the noninduced genes did not lack induction because they were repressed in T cells, but rather that expression was high and induction by IL-12 might have a more limited effect. To determine whether IL-12 was having effects on the noninduced genes we compared histone modifications at the promoters of three sustained genes and four noninduced genes. We observed that IL-12 stimulation increased H3 acetylation and decreased H3K27 methylation at both sets of genes, but had significantly greater FIGURE 5. STAT4 regulates TCR signaling. A, Heatmap of TCR sig- effects at the promoters of the induced genes (Fig. 6D). These data, by guest on September 26, 2021 naling-associated genes based on binding intensity of STAT4 to gene pro- coupled with the demonstration that STAT4 binds to the promoter Ϫ/Ϫ ϩ moters. B, Wild-type and Stat4 CD4 T cells were cultured under Th1 of Irf1 (Fig. 2), one of the noninduced genes, suggests that they are conditions for 5 days before RNA was isolated for qPCR using the indi- bona fide STAT4 targets, but that the effects of IL-12 are limited cated primers. RNA levels are relative to wild-type Th1 cells. C, Wild-type Ϫ Ϫ ϩ at these promoters for reasons that may include high basal expres- and Stat4 / CD4 T cells were cultured under Th1 conditions for 5 days before cells were stimulated with anti-CD3 for the indicated times. Total sion or lack of appropriate promoter context for induction. protein extracts were immunoblotted for phospho- and total ERK. D, Wild- Upon examining the genes that were in each group, we observed type and Stat4Ϫ/Ϫ CD4ϩ T cells were cultured under Th1 conditions for 5 some interesting patterns (Fig. 6E). Within the noninduced set days before cells were stimulated with anti-CD3 or PMA plus ionomycin were Foxp3 and Gzmb, genes not expressed in Th1 cells, and for 24 h. TNF-␣ was tested in cell-free supernatants using ELISA. Ms4a4b, a STAT4-independent Th1 gene. The transiently induced set included receptors such as Il2ra and Cd40l, and transcription factors such as Irf4, Irf8, and Bcl6 where transient expression Th1 expression (31). To avoid this problem, we analyzed might confer temporary function to the cell. The sustained subset Ϫ Ϫ TNF-␣ production from wild-type and Stat4 / Th1 cultures contained many of the genes that require STAT4 to program high stimulated with either anti-CD3 or PMA and ionomycin. Al- expression in Th1 cells including Ifng, Furin, Il18r1, and though TNF-␣ production following anti-CD3 stimulation is Gadd45g. We analyzed two additional genes that STAT4 is known Ϫ Ϫ decreased in Stat4 / cultures compared with wild-type, stim- to program but were not found on the ChIP-on-chip list, Etv5 and ulation with PMA and ionomycin, which bypasses the require- Hlx1, and found that they also were in this category. Thus, the ment for proximal signaling molecules such as SLP-76, resulted induction kinetics of STAT4 target genes following IL-12 stimu- in similar levels of TNF-␣ production (Fig. 5D). Thus, STAT4 lation of activated T cells is a strong predictor of the requirement targets genes associated with TCR signaling and contributes to for STAT4 in programming expression for Th1 cells. the potency of Ag receptor stimulation. Discussion Distinct patterns of STAT4-dependent gene induction among STAT4 is a critical regulator of adaptive inflammatory responses, STAT4 target genes yet there is still little known as to how it performs these functions. From the dataset of genes bound by STAT4, we wanted to deter- In previous work, we have shown that STAT4 binds to specific mine whether there was a relationship between the intensity of target genes and mediates programming for high levels of expres- STAT4 binding to a gene and the level of gene induction partic- sion in Th1 cells. It is still not well understood what genes are ularly because the overlap between our ChIP-on-chip targets and STAT4 targets on a genome-wide scale and how selectivity in the public domain expressed gene lists was limited (Fig. 1E). To activation is achieved. Previous microarray analyses have not pro- test this, we analyzed mRNA levels of over 30 genes, selected vided significant insight into gene programs activated by STAT4. The Journal of Immunology 3845 Downloaded from http://www.jimmunol.org/ by guest on September 26, 2021 FIGURE 6. Patterns of STAT4-dependent gene induction. A, Wild-type and Stat4Ϫ/Ϫ CD4ϩ T cells were activated for 3 days and stimulated with IL-12 for 4 or 18 h before RNA was isolated for qPCR. Graphs are plot of STAT4 binding intensity vs fold IL-12-stimulated gene induction at 4 h for genes that showed STAT4-dependent gene induction. B, Graph of STAT4 binding intensity vs fold-IL-12 induction of the genes identified as overlap between the ␤ ChIP-on-chip data and published microarray data (18) using fold-induction values from the microarray data. C, Graph of gene expression relative to 2m vs fold induction as determined in A. Genes that showed no induction or STAT4-dependent induction were divided into subsets based on the kinetics of expression (minimal induction, Յ 2-fold induction at 4 or 18 h; transient, Ͼ2-fold induction at 4 h, Յ 2-fold induction at 18 h; sustained induction, Ͼ2-fold induction at 4 and 18 h). D, ChIP assay for acetylated H3K9/18 or trimethylated H3K27 in purified CD4 T cells activated for three days and cultured with or without IL-12 for 4 h. qPCR for the promoters of induced genes (Furin, Ifng, IL18r1) and noninduced genes (Foxp3, Irf1, Mbd2, Ms4a4b) is expressed significantly different from induced genes (p Ͻ ,ء .as the fold-increase in acetylation or the percent decrease in H3K27 methylation from unstimulated cells 0.05) as determined by Student’s t test. E, Expression pattern of genes examined in A at 4 and 18 h after stimulation with IL-12 divided based on criteria in C. A partial list of gene names in each subset is indicated to the right of the graph.

In this report, we have used a ChIP-on-chip approach to identify 1B) would greatly skew subsequent analyses. For example, while STAT4 target genes and characterized STAT4 activity at these MEME identified a STAT-related site in 100% of the genes in our targets. curated list with a cut-off above 4, the algorithm identified Ͻ10% of One of the important issues in examining a large dataset such as genes with a similar motif when a list with a cut-off of three was used, the one in this report is the level of false positives and false neg- resulting from a large number of genes that lacked the site and de- atives identified. As we demonstrated in Fig. 3A, sequences from creased the enrichment potential for that motif. Regardless, this is a 60% of the genes in our curated list contained a consensus STAT limitation in the interpretation of this and related datasets. binding site, and 100% were identified with a match to a motif The Th1 genetic program involves many genes, which are either similar to a STAT consensus. This suggests a low rate of false- STAT4-dependent or STAT4-independent (9, 11). The 28 genes positives in the list, which is further confirmed by showing that we considered functionally linked to the Th1 phenotype are shown STAT4 binds to targets across a broad range of peak intensity in Table S3 and the STAT4 target genes in this dataset include values (Fig. 2). However, using the cut-off of peak value 4 for our cytokines (Ifng, Tnf, Lta), receptors (Il18r1, Il18rap, Il12rb2), and curated list might omit some valid STAT4 targets. When divided signaling factors (Gadd45b, Gadd45g, Myd88) (Fig. 7). STAT4 into ranges of peak values, the percent of sequences that contain a also bound Furin, which has critical roles in maintaining tolerance, STAT consensus decreases with decreasing peak value (Fig. 3B). In a function that suggests STAT4 might have both pro- and anti- the peak value 4.0–4.9 set the percentage is 50%, but decreases fur- inflammatory activities (32). It is interesting that 6 of the 20 known ther to 40% from 3.0–3.9 and 32% from 2.2–2.9, with an average of STAT4-dependent Th1 genes (Table S3), including Hlx1 and Etv5, 43% for the entire list. Thus, there might be STAT4 targets below our were not in our dataset. This likely represents either indirect acti- cut-off. However, the large number of genes in the lower ranges (Fig. vation of genes by STAT4, or STAT4 binding to target genes 3846 STAT4 TARGET GENES IN Th1 CELLS

might be compromised in Stat4Ϫ/Ϫ T cells, both from decreased growth in vitro, and the ability of strong stimuli, like anti-CD3 plus anti-CD28 or PMA plus ionomycin, to minimize the difference in IFN-␥ production between wild-type and Stat4Ϫ/Ϫ Th1 cells (30). Although we observed binding to a number of STAT4-bound TCR signaling genes in our dataset, we only identiﬁed one, Lcp2 en- coding SLP-76, that showed dependence on STAT4 for expression in Th1 cells. We did not observe decreased expression of Lcp2 in Stat4Ϫ/Ϫ naive T cells or Th2 cells (data not shown). As SLP-76 plays a central role in TCR signaling (29), multiple pathways could be affected. We demonstrated that anti-CD3-induced ERK activation was diminished in Stat4Ϫ/Ϫ Th1 cells. However, activation of downstream signaling molecules using PMA and ionomycin bypasses proximal signaling molecules and minimized the differences in cytokine production between wild-type and Stat4Ϫ/Ϫ Th1 cells. Thus, Stat4Ϫ/Ϫ T cells not only have a defect in the acqui- sition of the Th1 phenotype, but also in T cell activation of cells differentiated in a Th1-promoting environment.

FIGURE 7. Classiﬁcation of STAT4 target genes. STAT4 target genes Several important features of STAT4 binding vs gene regulation Downloaded from are clustered by function (receptors, cytokines) or functional subset (Th1, were also realized in this study including the lack of correlation TCR signaling). The Th1 cluster contains a subcluster of IL-18R-associ- between the intensity of binding and the fold-induction of gene ated genes and includes receptors and cytokines that are color-coded with expression. Although one might expect that more binding would other clusters. Genes are classiﬁed as STAT4-dependent, -independent, or result in greater transactivation, there are several reasons that may undetermined in relation to acute IL-12 restimulation. explain a disconnect between binding and gene induction. First,

the context of the promoter may dictate activity, demonstrating http://www.jimmunol.org/ outside of the promoter regions analyzed in this study, the latter that even in the context of a consensus-binding site, binding is not highlighting the limitation of a promoter- vs a genome-wide ChIP- the sole predictor of biological function. The focused time frame of on-chip. Regardless, Etv5 and Hlx1 are clearly STAT4-dependent our analysis might prevent observing induction of genes that re- Th1 genes with induction kinetics similar to other STAT4-depen- quire additional cofactors that were not present in IL-12-stimulated dent genes identified in the array (Fig. 6) (9). Conversely, 2 of the activated T cells. There may also be tissue-specific factors required 18 STAT4-bound Th1 genes (Fig. 4A) are not dependent on for STAT4 function to regulate these genes in other cell types. STAT4 for expression in Th1 cells, at least under standard in vitro Although no function for STAT4 in testis has been determined, we differentiation conditions. The Th1 lineage-promoting factor T-bet found 113 genes enriched in a testis microarray shared with the is also partially dependent upon STAT4 (9), though Tbx21 was not STAT4-bound dataset (Table S2C) that are unlikely to be induced by guest on September 26, 2021 identified in the STAT4 ChIP-on-chip dataset. As we recently in T cells. Second, the basal transcription of the gene may deter- demonstrated, many of the genes in the Th1 program are depen- mine the sensitivity of the gene to induction. We demonstrated that dent on both STAT4 and T-bet, including at least 7 of the 18 of the genes we tested where IL-12 had minimal induction, all had STAT4-bound Th1 genes (Fig. 4A and Table S3). The T-bet-de- high expression, compared with the control gene (Fig. 6C). Fi- pendence of all the genes in the STAT4-bound set has not been nally, STAT4 may bind to redundant consensus sites that mediate determined but these preliminary data suggest that there is at least gene activation by other STAT proteins. For example, Irf1 was a 30% overlap in the target genes of these two factors. bound by STAT4 but was not induced by IL-12. In contrast, IFN- Although STAT4 is clearly required for the establishment of the ␥/STAT1 induce Irf1 in a number of cell types. Similar results Th1 genetic program, how it mediates programming is still un- were observed for T-bet targets, where T-bet was found to bind but clear. STAT4 binds to target loci including Ifng, Hlx1, and Il18r1, not regulate gene targets (33). and mediates recruitment of Brg1-containing complexes, induces The benefit of genome-scale experiments is the opportunity to histone acetylation, and decreases histone methylation and the as- identify clusters of genes involved in a biological process within sociation of repressive enzymes such as DNMT3a and EZH2 (6, the activated gene regulatory network that contribute to specific 8–10). All of these events contribute to increased gene transcrip- cellular functions or phenotypes. Apart from the Th1 cluster, we tion and programming. In this report we identified distinct sets of identified a TCR signaling gene (Lcp2) that STAT4 programs for genes that are either transiently induced or have sustained induc- increased Ag receptor sensitivity and an IL-18 response cluster tion. Moreover, there was a correlation between sustained gene (Il18r1, Il18rap, Gadd45b, Gadd45g, Myd88) (Fig. 7) that sup- induction and programmed expression in Th1 cells. We have pre- ports previous observations on the requirement for STAT4 in viously analyzed an example of each of these gene categories; IL-18 sensitivity (28). A large number of IFN-induced genes were Il2ra is transiently induced, while Il18r1 is programmed for ex- also identified (Fig. 7), though the genes examined were not in- pression in Th1 cells (8, 23). At this point it is not clear whether duced to the same levels reported for IFN-␥ stimulation in other there are specific chromatin modifications that distinguish transient cell types (34). Recently, two groups have analyzed high-through- from sustained induction. However, having established many of put data from STAT1 ChIP (14, 35). Although these experiments the STAT4-dependent events that occur during Th1 differentiation were done in cultured cells (HeLa S3 or NIH3T3), and STAT1 has (9), and finding a set of genes, including Il2ra, that is only tran- differing spatial and temporal specificities of expression from siently induced, it will be possible to determine specific factors or STAT4, the proportion of chromatin immunoprecipitated STAT1 histone modifications that distinguish the effects of STAT4 at var- target sequences that contain a match to the consensus-binding ious gene promoters. sequences was similar to our studies. One of the subsets of genes identified in this study was involved In this report, using a combination of in silico and in vivo tech- in TCR signaling. Earlier reports had suggested that TCR signaling niques, we have identified STAT4 target genes in primary T cells. The Journal of Immunology 3847

Subsets of STAT4 targets mediate specific functions in Th1 cells. 15. Mathur, A. N., H. C. Chang, D. G. Zisoulis, R. Kapur, M. L. Belladonna, Analysis suggests that STAT4 target genes programmed for Th1 G. S. Kansas, and M. H. Kaplan. 2006. T-bet is a critical determinant in the instability of the IL-17-secreting T-helper phenotype. Blood 108: 1595–1601. expression have distinct activation kinetics, suggesting that it may 16. Labhart, P., S. Karmakar, E. M. Salicru, B. S. Egan, V. Alexiadis, B. W. be possible, using bioinformatics analysis, to identify genes that O’Malley, and C. L. Smith. 2005. Identification of target genes in breast cancer cells directly regulated by the SRC-3/AIB1 coactivator. Proc. Natl. Acad. Sci. play a role in inflammatory cell development and function. USA 102: 1339–1344. 17. Rogge, L., E. Bianchi, M. Biffi, E. Bono, S. Y. Chang, H. Alexander, C. Santini, G. Ferrari, L. Sinigaglia, M. Seiler, et al. 2000. Transcript imaging of the devel- Acknowledgments opment of human T helper cells using oligonucleotide arrays. Nat. Genet. 25: We thank C. Short for technical assistance and H. Nakshatri, C. Y. Lin, and 96–101. P. Labhart for review of this manuscript. 18. Watford, W. T., B. D. Hissong, L. R. Durant, H. Yamane, L. M. Muul, Y. Kanno, C. M. Tato, H. L. Ramos, A. E. Berger, L. Mielke, et al. 2008. Tpl2 kinase regulates T cell interferon-␥ production and host resistance to Toxoplasma gon- Disclosures dii. J. Exp. Med. 205: 2803–2812. 19. Bailey, T. L., and C. Elkan. 1994. Fitting a mixture model by expectation max- The authors have no financial conflict of interest. imization to discover motifs in biopolymers. Proc. Int. Conf. Intell. Syst. Mol. Biol. 2: 28–36. References 20. Crooks, G. E., G. Hon, J. M. Chandonia, and S. E. Brenner. 2004. WebLogo: a sequence logo generator. Genome Res. 14: 1188–1190. 1. Kaplan, M. H. 2005. STAT4: a critical regulator of inflammation in vivo. Im- 21. Beissbarth, T., and T. P. Speed. 2004. GOstat: find statistically overrepresented munol. Res. 32: 231–241. Gene Ontologies within a group of genes. Bioinformatics 20: 1464–1465. 2. Murphy, K. M., and S. L. Reiner. 2002. The lineage decisions of helper T cells. 22. Szabo, S. J., A. S. Dighe, U. Gubler, and K. M. Murphy. 1997. Regulation of the Nat. Rev. Immunol. 2: 933–944. ␤ interleukin (IL)-12R 2 subunit expression in developing T helper 1 (Th1) and 3. Robertson, M. J., H. C. Chang, D. Pelloso, and M. H. Kaplan. 2005. Impaired Th2 cells. J. Exp. Med. 185: 817–824.

␥ Downloaded from interferon- production as a consequence of STAT4 deficiency after autologous 23. O’Sullivan, A., H. C. Chang, Q. Yu, and M. H. Kaplan. 2004. STAT4 is required hematopoietic stem cell transplantation for lymphoma. Blood 106: 963–970. for interleukin-12-induced chromatin remodeling of the CD25 locus. J. Biol. 4. Remmers, E. F., R. M. Plenge, A. T. Lee, R. R. Graham, G. Hom, T. W. Behrens, Chem. 279: 7339–7345. P. I. de Bakker, J. M. Le, H. S. Lee, F. Batliwalla, et al. 2007. STAT4 and the risk 24. Xu, X., Y.-L. Sun, and T. Hoey. 1996. Cooperative DNA binding and sequence- of rheumatoid arthritis and systemic lupus erythematosus. N. Engl. J. Med. 357: selective recognition conferred by the STAT amino-terminal domain. Science 977–986. 273: 794–797. 5. Nguyen, K. B., W. T. Watford, R. Salomon, S. R. Hofmann, G. C. Pien, 25. Sandelin, A., W. Alkema, P. Engstrom, W. W. Wasserman, and B. Lenhard. A. Morinobu, M. Gadina, J. J. O’Shea, and C. A. Biron. 2002. Critical role for 2004. JASPAR: an open-access database for eukaryotic transcription factor bind- STAT4 activation by type 1 interferons in the interferon-␥ response to viral ing profiles. Nucleic Acids Res. 32: D91–94. http://www.jimmunol.org/ infection. Science 297: 2063–2066. 26. Pesu, M., L. Muul, Y. Kanno, and J. J. O’Shea. 2006. Proprotein convertase furin 6. Zhang, F., and M. Boothby. 2006. T helper type 1-specific Brg1 recruitment and is preferentially expressed in T helper 1 cells and regulates interferon ␥. Blood remodeling of nucleosomes positioned at the IFN-␥ promoter are Stat4 depen- 108: 983–985. dent. J. Exp. Med. 203: 1493–1505. 27. Lu, B., H. Yu, C. Chow, B. Li, W. Zheng, R. J. Davis, and R. A. Flavell. 2001. 7. Letimier, F. A., N. Passini, S. Gasparian, E. Bianchi, and L. Rogge. 2007. Chro- GADD45␥ mediates the activation of the p38 and JNK MAP kinase pathways matin remodeling by the SWI/SNF-like BAF complex and STAT4 activation and cytokine production in effector TH1 cells. Immunity 14: 583–590. synergistically induce IL-12R␤2 expression during human Th1 cell differentia- 28. Lawless, V. A., S. Zhang, O. N. Ozes, H. A. Bruns, I. Oldham, T. Hoey, tion. EMBO J. 26: 1292–1302. M. J. Grusby, and M. H. Kaplan. 2000. Stat4 regulates multiple components of 8. Yu, Q., V. T. Thieu, and M. H. Kaplan. 2007. Stat4 limits DNA methyltransferase IFN-␥-inducing signaling pathways. J. Immunol. 165: 6803–6808. recruitment and DNA methylation of the IL-18R␣ gene during Th1 differentia- 29. Smith-Garvin, J. E., G. A. Koretzky, and M. S. Jordan. 2009. T cell activation. tion. EMBO J. 26: 2052–2060. Annu. Rev. Immunol. 27: 591–619.

9. Thieu, V. T., Q. Yu, H. C. Chang, N. Yeh, E. T. Nguyen, S. Sehra, and 30. Afkarian, M., J. R. Sedy, J. Yang, N. G. Jacobson, N. Cereb, S. Y. Yang, by guest on September 26, 2021 M. H. Kaplan. 2008. Signal transducer and activator of transcription 4 is required T. L. Murphy, and K. M. Murphy. 2002. T-bet is a STAT1-induced regulator of for the transcription factor T-bet to promote T helper 1 cell-fate determination. IL-12R expression in naive CD4ϩ T cells. Nat. Immunol. 3: 549–557. Immunity 29: 679–690. 31. Chang, S., and T. M. Aune. 2005. Histone hyperacetylated domains across the 10. Yu, Q., H. C. Chang, A. N. Ahyi, and M. H. Kaplan. 2008. Transcription factor- IFN␥ gene region in natural killer cells and T cells. Proc. Natl. Acad. Sci. USA dependent chromatin remodeling of Il18r1 during Th1 and Th2 differentiation. 102: 17095–17100. J. Immunol. 181: 3346–3352. 32. Pesu, M., W. T. Watford, L. Wei, L. Xu, I. Fuss, W. Strober, J. Andersson, 11. Hoey, T., S. Zhang, N. Schmidt, Q. Yu, S. Ramchandani, X. Xu, L. K. Naeger, E. M. Shevach, M. Quezado, N. Bouladoux, et al. 2008. T-cell-expressed pro- Y. L. Sun, and M. H. Kaplan. 2003. Distinct requirements for the naturally oc- protein convertase furin is essential for maintenance of peripheral immune tol- curring splice forms Stat4␣ and Stat4␤ in IL-12 responses. EMBO J. 22: erance. Nature 455: 246–250. 4237–4248. 33. Beima, K. M., M. M. Miazgowicz, M. D. Lewis, P. S. Yan, T. H. Huang, and 12. Lund, R. J., Z. Chen, J. Scheinin, and R. Lahesmaa. 2004. Early target genes of A. S. Weinmann. 2006. T-bet binding to newly identiﬁed target gene promoters IL-12 and STAT4 signaling in Th cells. J. Immunol. 172: 6775–6782. is cell type-independent but results in variable context-dependent functional ef- 13. Ren, B., F. Robert, J. J. Wyrick, O. Aparicio, E. G. Jennings, I. Simon, fects. J. Biol. Chem. 281: 11992–12000. J. Zeitlinger, J. Schreiber, N. Hannett, E. Kanin, et al. 2000. Genome-wide lo- 34. Ramana, C. V., M. P. Gil, R. D. Schreiber, and G. R. Stark. 2002. Stat1-depen- cation and function of DNA binding proteins. Science 290: 2306–2309. dent and -independent pathways in IFN-␥-dependent signaling. Trends Immunol. 14. Robertson, G., M. Hirst, M. Bainbridge, M. Bilenky, Y. Zhao, T. Zeng, 23: 96–101. G. Euskirchen, B. Bernier, R. Varhol, A. Delaney, et al. 2007. Genome-wide 35. Hartman, S. E., P. Bertone, A. K. Nath, T. E. Royce, M. Gerstein, S. Weissman, proﬁles of STAT1 DNA association using chromatin immunoprecipitation and and M. Snyder. 2005. Global changes in STAT target selection and transcription massively parallel sequencing. Nat. Methods 4: 651–657. regulation upon interferon treatments. Genes Dev. 19: 2953–2968.