A chromatin-modifying function of JNK during stem cell differentiation

94 VOLUME 44 | NUMBER 1 | JANUARY 2012 Nature GeNetics

Signalingmediatescellularresponsestoextracellularstimuli.Thec-JunNH2-terminalkinase(JNK)pathwayexemplifiesonesubgroupofthemitogen-activatedprotein(MAP)kinases,which,besideshavingestablishedfunctionsinstressresponse,alsocontributetodevelopmentbyanunknownmechanism1–4.Weshowbygenome-widelocationanalysisthatJNKbindstoalargesetofactivepromotersduringthedifferentiationofstemcellsintoneurons.JNK-boundpromotersareenrichedwithbindingmotifsforthetranscriptionfactorNF-YbutnotforAP-1.NF-Yoccupiesthesepredictedsites,andoverexpressionofdominant-negativeNF-YAreducestheJNKpresenceonchromatin.WefindthathistoneH3Ser10(H3S10)isasubstrateforJNK,andJNK-boundpromotersareenrichedforH3S10phosphorylation.InhibitionofJNKsignalinginpost-mitoticneuronsreducesphosphorylationatH3S10andtheexpressionoftargetgenes.TheseresultsestablishMAPkinasebindingandfunctiononchromatinatanovelclassoftargetgenesduringstemcelldifferentiation.

Cell signaling provides the means to respond to various kinds of extra-cellular stimuli. This regulatory process entails intracellular pathways, which mediate changes in cellular physiology or transcriptional pro-grams through signaling cascades that activate downstream effectors. MAPK activation represents a paradigm of cellular signaling. JNK, a subgroup of the MAP kinases, is primarily activated by cytokines and exposure to environmental stress such as UV irradiation1–4. A number of findings support the idea that MAP kinases bring about transcriptional changes exclusively through their downstream effec-tors4,5. The AP-1 transcription factor is a prominent example of such an effector and is activated mostly by the phosphorylation of the c-Jun subunit and, in part, by the phosphorylation of related factors1–4. Notably, activated ERK1, ERK2 and p38 MAP kinases, as well as other signal-activated kinases, have been suggested to bind to chromatin at certain genes6–11, hinting at the possibility that MAP kinases could also act directly on gene promoters. Mouse embryonic stem (ES) cells do not require MAP kinase family members, including ERK, p38 or JNK, for self-renewal or the maintenance of pluripotency12. The knockout of individual JNK genes produces no detectable phe-notype in mice; however, mice lacking both JNK1 and JNK2 die at

mid-gestation, pointing to a developmental function for these kinases in addition to their established role in stress responses12. Moreover, ES cells lacking JNK1 show transcriptional deregulation of several lineage-commitment genes13 and fail to undergo neuronal differen-tiation, as do ES cells lacking JNK pathway scaffold proteins14. These findings argue for a developmental role for the JNK pathway, but the underlying mechanism remains unclear.

Here, we investigate JNK function in a well-established model of neuronal differentiation, in which pluripotent mouse ES cells progress to committed neuronal progenitors (NPs) and further to terminally differentiated pyramidal glutamatergic neurons (TNs)15–17. During ES cell differentiation, the mRNA levels of Mapk9 (encoding JNK2) stay constant, whereas those of Mapk8 (encoding JNK1) and Mapk10 (encoding JNK3) are elevated in TNs (Fig. 1a). A similar increase was observed at the protein level using an antibody that recognizes both JNK1 and JNK3 (JNK1/3) (Fig. 1b).

Notably, phosphorylation of JNK is only detected in TNs (Fig. 1c), suggesting a potential role for JNK signaling in fully differentiated cells. In addition, a large fraction of JNK proteins are localized to the nucleus in TNs, as previously shown in other instances of JNK activation (Fig. 1d)4,18,19. These observations, together with previous findings that MAP kinases may directly interact with chromatin6–11, led us to explore whether JNK binds to chromatin during neuronal differentiation. We performed chromatin immunoprecipitation using the JNK1/3-specific antibody followed by high-throughput sequenc-ing (ChIP-Seq). This analysis revealed that JNK binding is markedly enriched at many genomic sites in all three stem cell differentiation stages, with a strong preference for gene promoters (Fig. 1e,f), and most of the binding occurs in a small window comprising nucleotides −500 to +200 around the transcription start site (TSS) of genes (Fig. 1g and Supplementary Fig. 1a,b).

For further validation of the interaction of JNK with chromatin, we selected several JNK1/3 targets with different levels of enrichment and performed independent ChIP analysis coupled with quantitative PCR (ChIP-qPCR), which in each case confirmed the sequencing results (Fig. 2a). Notably, JNK1/3 enrichment increases during the course of differentiation to terminal neurons (Figs. 1e and 2a), a finding confirmed by a global analysis of the enrichment of JNK1/3 binding at TSSs (Supplementary Fig. 2). Of note, increased binding of JNK

A chromatin-modifying function of JNK during stem cell differentiationVijay K Tiwari1, Michael B Stadler1,2, Christiane Wirbelauer1, Renato Paro3,4, Dirk Schübeler1,4 & Christian Beisel3

1Friedrich Miescher Institute for Biomedical Research, Basel, Switzerland. 2Swiss Institute of Bioinformatics, Basel, Switzerland. 3Department of Biosystems Science and Engineering, Eidgenössische Technische Hochschule (ETH) Zürich, Basel, Switzerland. 4Faculty of Science, University of Basel, Basel, Switzerland. Corresponding should be addressed to D.S. ([email protected]) or M.B.S. ([email protected]).

Received 12 July; accepted 15 November; published online 18 December 2011; doi:10.1038/ng.1036

l e t t e r s©

201

2 N

atu

re A

mer

ica,

Inc.

All

rig

hts

res

erve

d.

Nature GeNetics VOLUME 44 | NUMBER 1 | JANUARY 2012 95

coincided with increased JNK1/3 expression, as well as an elevated presence of the active, phosphorylated form of JNK (Fig. 1a–c). To test whether JNK binding can also be detected in primary cells, we performed JNK1/3 ChIP assays on chromatin from the whole brain of 3.5-week-old adult mice. We tested select JNK targets by PCR, which in each case were found to be occupied by JNK in brain tissue, resembling what was seen in the neurons differentiated in vitro (Fig. 2b) and supporting the notion that JNK binding to chromatin is a general feature of neurons.

To gain insight into the biological role of JNK binding, we asked whether specific gene functions are overrepresented among the JNK target genes. Ingenuity Pathway Analysis (IPA) revealed a prominent enrichment for genes connected to developmental proc-esses (Table 1 and Supplementary Fig. 3). Of note, the most sig-nificant terms in TNs describing molecular and cellular functions included gene expression, cellular compromise, post-translational

modification and cell death (Table 1), which are functional catego-ries characteristic of canonical JNK signaling biology20.

Next, we looked for overlap between the location of genome-wide JNK targets and the presence of several features that define gene activity (RNA polymerase II (RNA Pol II) occupancy and mRNA abundance), open chromatin (histone H3 lysine 4 dimethylation (H3K4me2)) and repressed chromatin (histone H3 lysine 27 trimethy-lation (H3K27me3)). JNK-bound promoters have high RNA Pol II occupancy and an elevated occurrence of the H3K4me2 chromatin mark (Fig. 2c, Supplementary Figs. 1c,d and 2). Indeed, the major-ity of JNK target genes are expressed at a given differentiation stage, and levels of the repressive H3K27me3 chromatin modification are lower than the levels observed for genes negative for JNK binding (Fig. 2c, Supplementary Figs. 1c,d and 2). These observations were validated at single gene promoters (Supplementary Figs. 2,4 and data not shown). We conclude that genes bound by JNK1/3 are actively transcribed and are maintained in an open chromatin state. Notably, JNK-associated promoters represent only a subset of active genes and only a subset of those labeled by H3K4me2, suggesting

0123456

Mapk8 Mapk9 Mapk10

mR

NA

abu

ndan

ceno

rmal

ized

to G

apd

h ESNPTN JNK1/3

Lamin B1

p-JNK

Lamin B1

ES NP TN

Overlay

050050

0G3bp1

50ES

NP

TN

JNK1/3DAPI

ES NP TN

Promoter

ES NP TN

Fol

d ch

ange

rel

ativ

eto

who

le g

enom

e

1/21248

1632

−1,000 −500 0 500 1,0000

1

2

3

4

5

Position relative to TSS (nt)

Cov

erag

e(n

umbe

r of

rea

ds)

90−100%80−90%70−80%60−70%50−60%40−50%30−40%20−30%10−20%

Exon Intron Intergenic

a

d e

f g

b c Figure 1 JNK is upregulated during stem cell differentiation and directly binds promoters. (a) Analysis of Mapk8, Mapk9 and Mapk10 transcript levels by real-time PCR during neuronal differentiation of stem cells, revealing that Mapk8 and Mapk10 levels are low in ES cells but are upregulated upon differentiation. mRNA levels are plotted on the y axis and are normalized to Gapdh levels. Error bars represent s.e.m. (b) Protein blot detection of JNK1/3 in protein extracts isolated from ES, NP and TN cells showing protein upregulation during differentiation. Lamin B1 serves as loading control. (c) Protein blot analysis of samples in b using an antibody specific to phosphorylated JNK (p-JNK), revealing the presence of the active form in TNs. (d) Immunofluorescence with DAPI staining of the nucleus (blue) and an antibody specific to JNK1/3 (red). An overlay of DAPI and JNK staining shows that a substantial fraction of JNK localizes to the nucleus. Scale bar, 80 µm. (e) Genome browser view of enrichment for JNK binding at the G3bp1 locus in the three stages of ES cell differentiation as determined by ChIP. Tag densities are normalized to the total number of reads in each sample. (f) Peaks of JNK binding are enriched at promoters but not in exons, introns or intergenic regions in all three stages of ES cell differentiation. Enrichments are calculated relative to the genomic size of each type of region. (g) Averaged JNK coverage around TSSs in neurons normalized to the input chromatin control sample and grouped into ten classes of decreasing JNK1/3 ChIP-Seq signal.

Figure 2 JNK target promoters are active and show increased JNK binding during terminal differentiation. (a) ChIP-qPCR validation of the enrichment of JNK binding at various identified targets and non-target controls using JNK1/3 ChIP samples derived from all three stages of ES cell differentiation. Average enrichments from separate assays are plotted on the y axis as the ratio of precipitated DNA relative to the total input DNA and are further normalized to a control region. Error bars show s.e.m. (b) ChIP-qPCR validation of select JNK targets in JNK1/3 ChIP samples derived from adult mouse brain plotted as in a, showing the in vivo targeting of chromatin by JNK. (c) Genes were classified as JNK positive (enrichment at TSS ≥ 0.7, solid lines) or negative (enrichment at TSS < 0.7, dotted lines) on the basis of JNK levels at the TSS of each gene. The distribution of signals in TN cells for each group of genes is shown for RNA Pol II binding, mRNA levels and H3K4me2 and H3K27me3 chromatin modifications.

0

20

40

60

80

100

120

140

Cbx7Cbx5

Pax6

Frap1

Mcm

10Ta

nkTip

r1Tr

af2

Ppp2r1a

G3bp1

Pabpn1

Tob1Btg

1 In

g1

Fol

d en

richm

ent a

bove

ba

ckgr

ound ES

NP

TN

a b

c Non-targets JNK1/3 targets

0

2

4

6

8

10

Frap1

Mcm

10Ta

nkTr

af2

Ppp2r1a

Tob1

Btg1In

g1

Meg

f9Fol

d en

richm

ent a

bove

back

grou

nd

0 1 2

RNA Pol ll mRNA H3K4me2 H3K27me3

3 4

00.20.40.6

Den

sity 0.8

1.01.2

0 5 10 15

0

0.10

0.20

−1 0 1 2 3

0

0.2

0.4

0.6

0.8

JNK1/3PositiveNegative

JNK1/3PositiveNegative

JNK1/3PositiveNegative

JNK1/3PositiveNegative

−1 0 1 2 3 4

0

0.2

0.4

0.6

0.8

1.0

l e t t e r s©

201

2 N

atu

re A

mer

ica,

Inc.

All

rig

hts

res

erve

d.

96 VOLUME 44 | NUMBER 1 | JANUARY 2012 Nature GeNetics

l e t t e r s

that chromatin structure or gene activity alone cannot explain the specific chromatin localization of JNK.

To determine whether genomic targeting of JNK occurs by a genetic mechanism involving transcription factors, we asked whether occupied sites are enriched for certain DNA motifs. We analyzed the frequencies of heptamer sequence words in the top 1,000 JNK-bound regions in TNs (Supplementary Fig. 5a,b) and additionally used MEME21 to find de novo motifs (data not shown). Both analy-ses did not identify AP-1 binding sites as being overrepresented in JNK target genes (Supplementary Fig. 5a–c and data not shown). This is in spite of the observation that many genes responsive to JNK signaling harbor AP-1 consensus sequences in their promoter regions and are known targets of transcription factors, such as the Jun and Fos dimers that function in a JNK-dependent fashion2,22. Thus, direct binding of JNK to chromatin does not generally occur in the context of AP-1 sites. Of note, however, both sequence analysis methods identified two other prominent motifs: a highly enriched motif corresponding to the consensus binding site of NF-Y and a less enriched SP-1–like motif (Fig. 3a and Supplementary Fig. 5a). Comparison of the sequence of the NF-Y motif derived from the JNK ChIP-Seq data with a published NF-Y matrix23 revealed a close cor-respondence. A frequency profile of hits to the NF-YA–like sequence motif identified from JNK1/3-bound regions closely resembles the profile of JNK occupancy observed from the ChIP-Seq experiments (Supplementary Fig. 6). Moreover, when applying a linear model to predict JNK binding as a function of chromatin features and sequence motifs, the NF-Y motif holds stronger predictive power than the SP-1–like motif (Fig. 3b). At all three differentiation stages, the number of NF-Y sites combined with RNA Pol II occupancy can explain JNK binding almost as well as the full model that takes into account all available chromatin and sequence features (Fig. 3b and Supplementary Fig. 7a). In agreement with its lack of enrich-ment among JNK-bound sites, the AP-1 motif is not predictive of JNK binding, further confirming that JNK genomic binding does not occur in the context of the canonical AP-1 pathway (Fig. 3b and Supplementary Fig. 7a). The strong enrichment and predictive power of the NF-Y binding motif led us to explore NF-Y as a potential mediator of JNK recruitment to chromatin.

NF-Y is a trimeric complex in which NF-YA is the DNA-binding subunit. To test whether NF-Y indeed occupies JNK target gene

promoters, we performed ChIP-Seq using an antibody specific to NF-YA, which revealed that JNK and NF-YA exhibit the same binding location and behavior with increased signal strength in TNs (Fig. 3c,d and Supplementary Figs. 2 and 7b). A global comparison of sites in promoters enriched for JNK and NF-YA binding confirmed that both bind to the same genomic sites, as is evident from a high cor-relation at all three stages of stem cell differentiation (Fig. 3e and Supplementary Figs. 2 and 7c). Addition of the experimental data on NF-YA binding to the linear model further increased its perform-ance (Fig. 3b and Supplementary Fig. 7a). We conclude that NF-Y occupies the same chromosomal sites as JNK1/3.

We investigated the relationship between NF-Y and JNK through the use of HEK293 cells, which are more amendable to overexpression studies and show similar localization for both proteins to that seen in TNs (Supplementary Fig. 8). To explore whether NF-Y is a poten-tial mediator of JNK recruitment to chromatin, we overexpressed an NF-YA mutant protein previously shown to act as a dominant-negative repressor of endogenous NF-YA24. Overexpression of mutant NF-YA led to a significant reduction in JNK binding at target gene promoters (Fig. 3f). Together with the observation that NF-YA binding itself is unaffected in JNK-deficient fibroblasts (Supplementary Fig. 9), this finding suggests a critical role for NF-Y in mediating JNK presence at promoters.

To determine whether JNK binding and its potential function on chromatin involve its kinase activity, we exposed cells to SP600125, an established inhibitor of JNK signaling25. Exposure of ES cells to this compound neither reduced pluripotency, as measured by particular markers, nor the ability to differentiate to the progenitor state (data not shown). This is in line with the absence of JNK phosphorylation (Fig. 1c) and previous observations that this pathway is inactive in stem cells12. However, when SP600125 was added during early stages of neuronal culture, cells died rapidly and neurogenesis was severely abrogated (Fig. 4a and Supplementary Fig. 10a), suggesting that JNK kinase activity is crucial for terminal neuronal differentiation in this in vitro system. Exposure of already differentiated neurons to SP600125 led to a significant decrease in JNK phosphorylation (Fig. 4b) but did not result in detectable apoptosis (data not shown).

We assessed whether JNK could modify chromatin components by surveying the phosphorylation state of several histone residues in the presence or absence of the JNK inhibitor in terminally differentiated, post-mitotic neurons. In particular, we screened established histone substrates for phosphorylation by signaling kinases26. This analysis revealed that JNK inhibition leads to a substantial reduction in phosphorylation of histone H3 at serine 10 (H3S10) (Fig. 4c). In agreement with the effect of the inhibitor in cells, in vitro kinase assays showed that H3S10 is a potential substrate for phosphoryla-tion by JNK (Fig. 4d). Consistent with these observations, ChIP assays revealed increased H3S10 phosphorylation levels specifically at JNK target genes (Fig. 4e).

Having established that JNK acts on chromatin, we next compared the transcriptome of neurons in the presence or absence of SP600125 to determine whether the inhibition of JNK signaling affects gene expression. This analysis revealed that JNK target genes are indeed preferentially downregulated in cells exposed to JNK inhibitor when compared to non-target genes expressed at similar levels (Fig. 5a). Because JNK and NF-Y bind to the same set of genes, a similar trend is observed for NF-Y target genes (Fig. 5b). Notably, such downregu-lation does not occur at highly expressed genes but is restricted to target genes expressed at moderate levels. Even though moderately expressed, these are clearly actively transcribed, as evidenced by more sensitive quantification using RNA-Seq (Supplementary Fig. 10b,c).

table 1 Molecular and cellular functions and physiological activities enriched for genes occupied by JNK in tN cellsName P value Number of molecules

Molecular and cellular functionsGene expression >4.33 × 10–10 222

Cellular compromise >1.22 × 10–8 116

Cell cycle >5.03 × 10–7 138

Post-translational modification >4.11 × 10–6 126

Cell death >8.03 × 10–6 251

Physiological system development and function

Embryonic development >2.00 × 10–7 102

Organismal development >2.38 × 10–3 99

Cardiovascular system development and function

>3.14 × 10–3 23

Nervous system development and function

>3.14 × 10–3 25

Organ development >3.14 × 10–3 14

Shown are the top five scoring hits in these categories using IPA, together with significance scores (P values) and the number of genes included in each class.

© 2

012

Nat

ure

Am

eric

a, In

c. A

ll ri

gh

ts r

eser

ved

.

Nature GeNetics VOLUME 44 | NUMBER 1 | JANUARY 2012 97

l e t t e r s

Decreased expression at select single JNK target genes (Mcm10 and Traf2) validated the reduction in gene expression (Fig. 5c).

Assessing whether changes in transcription are linked to JNK occupancy, we performed ChIP-Seq with the JNK1/3-specific anti-body in neurons exposed to DMSO control or JNK inhibitor. In this analysis, the majority of sites did not show a change in JNK binding, suggesting that JNK inhibition does not cause a general displacement of JNK from chromatin (Supplementary Fig. 10d,e). However, when comparing transcription with JNK binding, we observed that genes with reduced expression upon exposure to JNK inhibitor showed reduced binding of JNK (P < 2.22 × 10−16; Fig. 5d),

suggesting that reduction in the expression of target genes is directly linked to the reduced presence of JNK at their promoters.

Taken together, our data suggest a model of JNK recruitment to NF-Y–bound promoters during stem cell differentiation in order to phosphorylate histone H3 at Ser10 and modulate gene expres-sion. These results suggest a new pathway in mammalian cells, in which MAP kinases regulate gene expression by directly binding to gene promoters and phosphorylating chromatin components. Several lines of evidence support this model. JNK is present in the nucleus and binds to specific regions, which are co-occupied by the NF-Y transcription factor. This co-localization of JNK and NF-Y is

JNK1/3

NF

-YA

c

e3

f

0

2

4

6

8

10

12

14

16

TRAF2 MCM10 PER1

Fol

d en

richm

ent a

bove

back

grou

nd

WT NF-YAWT NF-YA +TetDN NF-YADN NF-YA +Tet

ES JNK1/3

NP JNK1/3

TN JNK1/3

ES NF-YA

NP NF-YA

TN NF-YA

G3bp1

50

050

050

015

015

015

0

2

1

0

–1

–1 0 1 2 3 4 5

R = 0.718

SP-1

aNF-Y2

1

0

Bits

2

1

0

Bits

1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9

b

Mod

el p

erfo

rman

ce (

adju

sted

r2 )

full

RNA Pol

II

H3K27

me3

H3K4m

e2

mRNA

NF-Y lik

e

SP-1 lik

e

AP-1 lik

e

RNA Pol

II+NF-Y

like

RNA Pol

II+SP-1

like

RNA Pol

II+AP-1

like

0

0.1

0.2

0.3

0.4

0.5

0.6

0

0.1

0.2

0.3

0.4

0.5

0.6

Mod

el p

erfo

rman

ce (

adju

sted

r2 )

full

full+

NF-YA

NF-YA

RNA Pol

II+NF-Y

A

RNA Pol

II+NF-Y

like+

NF-YA

0

10

20

30

40

50

60

70

80

Mcm

10Cbx7

Cbx5Pax

6

Frap1

Tank

Tipr1

Traf

2

Ppp2r1a

G3bp1

Pabpn1

Tob1

Btg1

Ing1

Fol

d en

richm

ent a

bove

back

grou

nd

ES

NP

TN

Non-targets JNK1/3 targets

d

Figure 3 NF-Y mediates JNK recruitment to chromatin. (a) Sequence logos of motifs identified in JNK target genes. (b) Performance (adjusted r2) of different linear models in predicting JNK binding at promoters in TNs as a function of chromatin features (RNA Pol II binding and H3K27me3 and H3K4me2 marks), sequence features (number of NF-Y–, SP-1– or AP-1–like motifs), mRNA levels, or a combination of all (full) or two (RNA Pol II binding and a given motif) of these features. RNA Pol II + NF-Y–like motif almost achieve the performance of the full model, explaining 40% of the observed variance in JNK binding (left). Models including the NF-YA binding data derived from ChIP-Seq (right) explain more than 60% of the variance. (c) Genome browser screenshot comparing JNK and NF-YA binding at the G3bp1 locus showing similar binding dynamics during neuronal differentiation. (d) NF-YA enrichment as determined using ChIP-qPCR for various JNK targets in ES, NP and TN stages. NF-Y occupies all tested JNK targets and shows similar binding dynamics during differentiation. Error bars show s.e.m. (e) Comparison of enrichment in JNK and NF-YA binding at promoters in TN cells. Promoters with proximal JNK peaks (within 1 kb of the TSS) are shown in red. Dotted lines indicate the cut-off values used in this study to define promoters positive for JNK and NF-YA binding (0.7 and 0.5, respectively). (f) Dominant-negative NF-YA reduces JNK binding to chromatin. HEK293 cells were stably transfected with constructs encoding wild-type (WT) or dominant-negative (DN) NF-YA, and NF-YA expression was induced by the addition of tetracycline (Tet) to cells. Subsequently, JNK binding was determined by ChIP-qPCR using primers specific for JNK target genes. Error bars represent s.e.m.

© 2

012

Nat

ure

Am

eric

a, In

c. A

ll ri

gh

ts r

eser

ved

.

98 VOLUME 44 | NUMBER 1 | JANUARY 2012 Nature GeNetics

l e t t e r s

likely functionally important, since overexpression of a dominant- negative NF-Y mutant that cannot bind DNA reduces JNK binding to chromatin. This dependency of JNK on NF-Y for chromatin binding is in line with the results of a previous study that implicated NF-Y in the regulation of the JNK pathway during Drosophila thorax development27 and is compatible with reports linking NF-Y and the MAPK pathways28–30. Of note, we show that JNK target promoters are preferentially enriched for phosphorylated H3S10. JNK phos-phorylates H3S10 in vitro, and inhibition of JNK phosphorylation

reduces phosphorylation at this site and diminishes target gene expression in neuronal cells. This activity of JNK has been identified in a validated in vitro model of neurogenesis using non-transformed primary cells and was confirmed in primary neurons, strongly argu-ing that it is a canonical activity of JNK during development. Our data also agree with observations that activated JNK is required for axon formation and dendritic growth31,32 and are supported by our observation that early neuronal states are sensitive to the inhibition of JNK signaling.

Figure 4 Inhibition of JNK signaling blocks differentiation and reduces H3S10 phosphorylation. (a) Exposure of cells to the JNK inhibitor SP600125 (JNKi) during the transition from neuronal progenitors to neurons abrogates neurogenesis. NPs were treated with either DMSO (control) or with SP600125 for 2 h after plating, and light-field microscopy images were taken 24 h later. Scale bar, 70 µm. (b) TNs were exposed to either DMSO or SP600125, and the levels of phosphorylated JNK were detected by protein blotting. The levels of total JNK1/3 and Lamin B1 (loading control) are also shown. Exposure to SP600125 leads to a substantial reduction in the levels of phosphorylated JNK. (c) Total histones were isolated from TNs exposed to either DMSO or SP600125, and levels of phosphorylated H3S10 (p-H3S10), H3T3 (p-H3T3), H3T11 (p-H3T3) and H3S28 (p-H3S28), as well as acetylated H3 (H3Ac), H3K4me2, H3K9me3 and total histone H4 were detected by protein blotting. Total histone H3 levels are also shown as a loading control. Inhibition of JNK signaling leads to a substantial reduction in the levels of phosphorylated H3S10. (d) JNK phosphorylates H3S10 in vitro. Recombinant active JNK was incubated with ATP and recombinant H3.1 and kinase reactions were performed, followed by protein blotting with the indicated antibodies. (e) Phosphorylated H3S10 is preferentially enriched at JNK target genes. ChIP-qPCR for phosphorylated H3S10 presence at various JNK target and non-target control genes using phosphorylated H3S10 ChIP samples derived from the TN cells. Error bars represent s.e.m.

a c

Exp

ress

ion

chan

ge (

log 2

JNK

i:DM

SO

)

dbJNK1/3 targets (N = 3,176)

Other genes (N = 11,898)

NF-YA targets (N = 1,297)

Other genes (N = 13,777)

0.5

0

–0.5

0.5

1.6

0

–0.5

Expression bins

* * * * * * * * *

(2.2

–3.3

]

(3.3

–3.9

]

(3.9

–4.5

]

(4.5

–5]

(5–5

.6]

(5.6

–6.2

]

(6.2

–6.8

]

(6.8

–7.6

]

(7.6

–8.5

]

(8.5

–14]

Expression bins

(2.2

–3.3

]

(3.3

–3.9

]

(3.9

–4.5

]

(4.5

–5]

(5–5

.6]

(5.6

–6.2

]

(6.2

–6.8

]

(6.8

–7.6

]

(7.6

–8.5

]

(8.5

–14]

Exp

ress

ion

chan

ge (

log 2

JNK

i:DM

SO

)

1.41.21.00.80.60.40.2

0

2.5

0.05

Rel

ativ

e m

RN

Ale

vels

Rel

ativ

e m

RN

Ale

vels

2.0

1.5

1.0

0.5

0Traf2

Mcm10

DMSOJNKi

Cha

nge

in J

NK

bin

ding

at T

SS

(JN

Ki –

DM

SO

)

0

� = 0.255 (P < 2.22e–16)

–0.05

–0.10

–0.15

–0.20

–0.25

–0.30

–0.4 –0.2

Expression change (JNKi – DMSO)

0 0.2 0.4

Figure 5 Blocking JNK kinase activity downregulates target gene expression. (a) Effect of JNK inhibition on the transcriptome. Genes were divided into ten equally sized bins according to expression levels, median-centered and SP600125-induced expression changes were compared between JNK target and non-target genes. Bins with significant differences in expression between the two gene classes are indicated by asterisks (P value < 0.001, two-sided Wilcoxon rank-sum test with continuity correction). These data show preferential downregulation of JNK targets with moderate expression levels. Dotted lines show the minimum and maximum range of data. (b) Analysis performed as in a for NF-YA target genes illustrating the same preferential downregulation by the inhibition of JNK kinase activity. (c) Real-time PCR analysis of transcript levels of two JNK targets (Mcm10 and Traf2) in TNs exposed to either DMSO or SP600125. Relative mRNA levels were determined by normalization to Gapdh expression, and average data from independent assays are plotted on the y axis. Error bars show s.e.m. (d) Correlation between expression changes caused by the inhibition of JNK kinase activity (x axis) and changes in JNK binding at the TSS of relevant genes (y axis). The red line corresponds to a local smoothed line fitted to the values for individual genes (loess fit with smoothing parameter α = 0.15). Changes in expression and JNK binding are significantly correlated (Spearman’s rank correlation ρ = 0.255, P < 2.2 × 10−16).

a

0

0.5

1.0

1.5

2.0

2.5

Gapdh

Hprt

Tmef

f2

Epha3

Epha4

Frap1

Mcm

10Per

1Per

2Per

3Tr

af2

Fol

d en

richm

ent a

bove

back

grou

nd

Non-targets JNK1/3 targets

p-JNK

JNK1/3

Lamin B1

DMSO JNKi

p-H3S10

H3

p-H3S10

H3

Auror

a B

No kin

ase

DMSO

JNKi p-H3T3

p-H3T11

p-H3S28

H3Ac

H3K4me2

H3K9me3

H4

DMSO

JNKi

c d

be

JNK2

JNK3

JNK1

© 2

012

Nat

ure

Am

eric

a, In

c. A

ll ri

gh

ts r

eser

ved

.

Nature GeNetics VOLUME 44 | NUMBER 1 | JANUARY 2012 99

l e t t e r s

Our study adds to several recent reports of signaling kinases that bind and act on chromatin. The mitogen- and stress-activated kinase 1 (MSK1)6 and the adenosine monophosphate–activated protein kinase (AMPK)10 were shown to be present on promoters for gene activation. Additionally, the JAK2 kinase was shown to phosphorylate histone H3 at tyrosine 41 (H3Y41), excluding HP1α from chromatin and thereby activating gene expression33. In yeast, it has been reported that not only stress-activated protein kinases (SAPKs) but also most MAP kinases become physically associated with response genes, and this interaction was determined to be essential for proper cellular adaptation to extracellular stimuli8,34.

H3S10 phosphorylation is a previously described hallmark of mitotic chromosomes, which is also known to associate with relaxed chromatin and active transcription in interphase26,35–37. High H3S10 phosphorylation levels block binding of the heterochromatic protein HP1 (ref. 38), and reduced levels of the kinase that phosphorylates this site in interphase, JIL-1, lead to the spreading of the major H3K9me2 heterochromatin marker and the binding of HP1 to ectopic locations35. These observations suggested that H3S10 phosphoryla-tion prevents ectopic recruitment or spreading of heterochromatic factors like HP1 and thereby stabilizes an active transcription state. Our data implicate JNK as the first MAP kinase that phosphorylates H3S10, although other kinases including PIM1, IKKα, MSK1 and MSK2, PKB/Akt and RSK2 have previously been reported to modify this site39–45. These studies established that kinase-mediated H3S10 phosphorylation accompanies proper gene expression in response to upstream signals39–45 in a paradigm that extends to JNK. We show that a mammalian MAP kinase, JNK, is recruited to a large set of gene promoters, presumably in response to differentiation signals, and that it facilitates the activation of target genes by directly modifying chro-matin components. This represents a new pathway by which MAP kinases regulate gene expression and greatly expands the number of target genes downstream of JNK activation. It is possible that compa-rable principles of chromatin binding and activity might be relevant for other MAP kinases.

URLs. Mouse genome assembly (July 2007), http://www.ncbi.nlm.nih.gov/genome/guide/mouse/; UCSC RefSeq transcript annotation (18 October 2009), http://hgdownload.cse.ucsc.edu/goldenPath/mm9/database/refGene.txt.gz; mouse microRNA sequences, ftp://ftp.sanger.ac.uk/pub/mirbase/sequences/13.0; GenBank mouse sequences for ribosomal RNA (rRNA), small nuclear RNA (snRNA), small nucleolar RNA (snoRNA) and RefSeq mRNA (16 July 2009), http://www.ncbi.nlm.nih.gov/sites/entrez/, mouse transfer RNA (tRNA) sequences, http://lowelab.ucsc.edu/GtRNAdb/; NCBI piwi-interacting RNA (piRNA) sequences (DQ539889–DQ569912), http://www.ncbi.nlm.nih.gov/nuccore/; R, http://www.r-project.org/.

MeTHodSMethods and any associated references are available in the online version of the paper at http://www.nature.com/naturegenetics/.

Accession numbers. All data have been deposited at the GEO database (GSE25533).

Note: Supplementary information is available on the Nature Genetics website.

ACKnoWleDgMenTSWe thank L. Hoerner, M. Seimiya and I. Nissen for technical assistance, N. Tiwari for help with reagents and critical suggestions, G. Christofori, M. Bentires-Alj and members of the Schübeler laboratory for comments to the manuscript, R. Davis (University of Massachusetts Medical School, Worcester, Massachusetts, USA) for providing wild-type NIH3T3 cells and those lacking JNK1 and JNK2,

and R. Mantovani (University of Milan, Milan, Italy) for the wild-type and dominant-negative NF-Y constructs. We thank the facilities at the Friedrich Miescher Institute, especially T. Roloff. Illumina sequencing was carried out at the Quantitative Genomics Facility of the Department of Biosystems Science and Engineering (D-BSSE), ETH Zurich. V.K.T. is supported by a Marie Curie International Incoming fellowship and an EMBO long-term postdoctoral fellowship. Research in the laboratory of R.P. is supported by the Deutsche Forschungsgemeinschaft (DFG), the European Union Epigenome Network of Excellence (EU-NoE) Epigenesys, Novartis and the ETH Zürich. Research in the laboratory of D.S. is supported by the Novartis Research Foundation, the European Union ((EU-NoE) Epigenesys, Blueprint)), the European Research Council (ERC-204264) and SystemsX.ch (Cell Plasticity).

AUTHoR ConTRIBUTIonSV.K.T. initiated and designed the study, performed experiments, analyzed data and wrote the manuscript. M.B.S. designed and performed the computational analysis and wrote the manuscript. C.W. performed experiments and analyzed data. R.P. provided input during the study and comments on the manuscript. C.B. initiated the study, performed experiments, analyzed data and wrote the manuscript. D.S. designed the study, analyzed data and wrote the manuscript.

CoMPeTIng FInAnCIAl InTeReSTSThe authors declare no competing financial interests.

Published online at http://www.nature.com/naturegenetics/. Reprints and permissions information is available online at http://www.nature.com/reprints/index.html.

1. Wagner, E.F. & Nebreda, A.R. Signal integration by JNK and p38 MAPK pathways in cancer development. Nat. Rev. Cancer 9, 537–549 (2009).

2. Karin, M. & Gallagher, E. From JNK to pay dirt: jun kinases, their biochemistry, physiology and clinical importance. IUBMB Life 57, 283–295 (2005).

3. Weston, C.R. & Davis, R.J. The JNK signal transduction pathway. Curr. Opin. Cell Biol. 19, 142–149 (2007).

4. Bogoyevitch, M.A. & Kobe, B. Uses for JNK: the many and varied substrates of the c-Jun N-terminal kinases. Microbiol. Mol. Biol. Rev. 70, 1061–1095 (2006).

5. Pearson, G. et al. Mitogen-activated protein (MAP) kinase pathways: regulation and physiological functions. Endocr. Rev. 22, 153–183 (2001).

6. Gehani, S.S. et al. Polycomb group protein displacement and gene activation through MSK-dependent H3K27me3S28 phosphorylation. Mol. Cell 39, 886–900 (2010).

7. Lawrence, M.C. et al. The roles of MAPKs in disease. Cell Res. 18, 436–442 (2008).

8. Pokholok, D.K., Zeitlinger, J., Hannett, N.M., Reynolds, D.B. & Young, R.A. Activated signal transduction kinases frequently occupy target genes. Science 313, 533–536 (2006).

9. Hu, S. et al. Profiling the human protein-DNA interactome reveals ERK2 as a transcriptional repressor of interferon signaling. Cell 139, 610–622 (2009).

10. Bungard, D. et al. Signaling kinase AMPK activates stress-promoted transcription via histone H2B phosphorylation. Science 329, 1201–1205 (2010).

11. Simone, C. et al. p38 pathway targets SWI-SNF chromatin-remodeling complex to muscle-specific loci. Nat. Genet. 36, 738–743 (2004).

12. Binétruy, B., Heasley, L., Bost, F., Caron, L. & Aouadi, M. Concise review: regulation of embryonic stem cell lineage commitment by mitogen-activated protein kinases. Stem Cells 25, 1090–1095 (2007).

13. Amura, C.R., Marek, L., Winn, R.A. & Heasley, L.E. Inhibited neurogenesis in JNK1-deficient embryonic stem cells. Mol. Cell. Biol. 25, 10791–10802 (2005).

14. Xu, P. et al. In vitro development of mouse embryonic stem cells lacking JNK/stress-activated protein kinase–associated protein 1 (JSAP1) scaffold protein revealed its requirement during early embryonic neurogenesis. J. Biol. Chem. 278, 48422–48433 (2003).

15. Plachta, N., Bibel, M., Tucker, K.L. & Barde, Y.A. Developmental potential of defined neural progenitors derived from mouse embryonic stem cells. Development 131, 5449–5456 (2004).

16. Bibel, M., Richter, J., Lacroix, E. & Barde, Y.A. Generation of a defined and uniform population of CNS progenitors and neurons from mouse embryonic stem cells. Nat. Protoc. 2, 1034–1043 (2007).

17. Bibel, M. et al. Differentiation of mouse embryonic stem cells into a defined neuronal lineage. Nat. Neurosci. 7, 1003–1009 (2004).

18. Raman, M., Chen, W. & Cobb, M.H. Differential regulation and properties of MAPKs. Oncogene 26, 3100–3112 (2007).

19. Aplin, A.E., Hogan, B.P., Tomeu, J. & Juliano, R.L. Cell adhesion differentially regulates the nucleocytoplasmic distribution of active MAP kinases. J. Cell Sci. 115, 2781–2790 (2002).

20. Weston, C.R. & Davis, R.J. The JNK signal transduction pathway. Curr. Opin. Genet. Dev. 12, 14–21 (2002).

21. Bailey, T.L., Williams, N., Misleh, C. & Li, W.W. MEME: discovering and analyzing DNA and protein sequence motifs. Nucleic Acids Res. 34, W369–W373 (2006).

22. Shaulian, E. AP-1—the Jun proteins: oncogenes or tumor suppressors in disguise? Cell. Signal. 22, 894–899 (2010).

© 2

012

Nat

ure

Am

eric

a, In

c. A

ll ri

gh

ts r

eser

ved

.

100 VOLUME 44 | NUMBER 1 | JANUARY 2012 Nature GeNetics

23. Dolfini, D., Zambelli, F., Pavesi, G. & Mantovani, R. A perspective of promoter architecture from the CCAAT box. Cell Cycle 8, 4127–4137 (2009).

24. Mantovani, R. et al. Dominant-negative analogs of NF-YA. J. Biol. Chem. 269, 20340–20346 (1994).

25. Bennett, B.L. et al. SP600125, an anthrapyrazolone inhibitor of Jun N-terminal kinase. Proc. Natl. Acad. Sci. USA 98, 13681–13686 (2001).

26. Baek, S.H. When signaling kinases meet histones and histone modifiers in the nucleus. Mol. Cell 42, 274–284 (2011).

27. Yoshioka, Y., Suyari, O. & Yamaguchi, M. Transcription factor NF-Y is involved in regulation of the JNK pathway during Drosophila thorax development. Genes Cells 13, 117–130 (2008).

28. Wood, J.L. & Russo, A.F. Autoregulation of cell-specific MAP kinase control of the tryptophan hydroxylase promoter. J. Biol. Chem. 276, 21262–21271 (2001).

29. Alabert, C. et al. Cell type-dependent control of NF-Y activity by TGF-β. Oncogene 25, 3387–3396 (2006).

30. Lecona, E. et al. Upregulation of annexin A1 expression by butyrate in human colon adenocarcinoma cells: role of p53, NF-Y, and p38 mitogen-activated protein kinase. Mol. Cell. Biol. 28, 4665–4674 (2008).

31. Rosso, S.B., Sussman, D., Wynshaw-Boris, A. & Salinas, P.C. Wnt signaling through Dishevelled, Rac and JNK regulates dendritic development. Nat. Neurosci. 8, 34–42 (2005).

32. Oliva, A.A. Jr. Atkins, C.M., Copenagle, L. & Banker, G.A. Activated c-Jun N-terminal kinase is required for axon formation. J. Neurosci. 26, 9462–9470 (2006).

33. Dawson, M.A. et al. JAK2 phosphorylates histone H3Y41 and excludes HP1α from chromatin. Nature 461, 819–822 (2009).

34. Edmunds, J.W. & Mahadevan, L.C. MAP kinases as structural adaptors and enzymatic activators in transcription complexes. J. Cell Sci. 117, 3715–3723 (2004).

35. Zhang, W. et al. The JIL-1 histone H3S10 kinase regulates dimethyl H3K9 modifications and heterochromatic spreading in Drosophila. Development 133, 229–235 (2006).

36. Prigent, C. & Dimitrov, S. Phosphorylation of serine 10 in histone H3, what for? J. Cell Sci. 116, 3677–3685 (2003).

37. Nowak, S.J. & Corces, V.G. Phosphorylation of histone H3: a balancing act between chromosome condensation and transcriptional activation. Trends Genet. 20, 214–220 (2004).

38. Fischle, W. et al. Regulation of HP1-chromatin binding by histone H3 methylation and phosphorylation. Nature 438, 1116–1122 (2005).

39. Anest, V. et al. A nucleosomal function for IκB kinase-α in NF-κB–dependent gene expression. Nature 423, 659–663 (2003).

40. Cerutti, H. & Casas-Mollano, J.A. Histone H3 phosphorylation: universal code or lineage specific dialects? Epigenetics 4, 71–75 (2009).

41. Dyson, M.H. et al. MAP kinase–mediated phosphorylation of distinct pools of histone H3 at S10 or S28 via mitogen- and stress-activated kinase 1/2. J. Cell Sci. 118, 2247–2259 (2005).

42. Pérez-Cadahía, B., Drobic, B. & Davie, J.R. H3 phosphorylation: dual role in mitosis and interphase. Biochem. Cell Biol. 87, 695–709 (2009).

43. Sassone-Corsi, P. et al. Requirement of Rsk-2 for epidermal growth factor–activated phosphorylation of histone H3. Science 285, 886–891 (1999).

44. Yamamoto, Y., Verma, U.N., Prajapati, S., Kwak, Y.T. & Gaynor, R.B. Histone H3 phosphorylation by IKK-α is critical for cytokine-induced gene expression. Nature 423, 655–659 (2003).

45. Zippo, A., De Robertis, A., Serafini, R. & Oliviero, S. PIM1-dependent phosphorylation of histone H3 at serine 10 is required for MYC-dependent transcriptional activation and oncogenic transformation. Nat. Cell Biol. 9, 932–944 (2007).

l e t t e r s©

201

2 N

atu

re A

mer

ica,

Inc.

All

rig

hts

res

erve

d.

Nature GeNeticsdoi:10.1038/ng.1036

oNLINeMeTHodSCell culture. ES cells were derived from a cross between 129Sv and C57Bl/6 mice and were cultured and differentiated as previously described16,46.

Protein blot analysis. SDS-PAGE was performed with total protein extracts, and rabbit polyclonal antibody recognizing JNK1 and JNK3 (JNK1/3), phos-phorylated JNK or Lamin B1 were used for protein blot analysis. Blots were developed with enhanced chemiluminescence (ECL) reagent (GE Healthcare). For the detection of histone modifications, acid extraction was performed according to the Upstate/Millipore protocol. Histones were separated by 15% SDS-PAGE, and blots were probed with rabbit polyclonal antibodies recog-nizing phosphorylated H3S10, H3T11 or H3S28, acetylated H3, total H4, H3K4me2 (all from Millipore), total H3, H3K9me3 (both from Abcam) or phosphorylated H3T3 (Cell Signaling Technology).

Immunofluorescence analysis. Cells were fixed with 4% PFA for 10 min and were rinsed in PBS and permeabilized with 0.2% Triton in PBS for 5 min. Blocking was performed in 3.5% goat serum for 15 min. After being washed in PBS, cells were incubated with a primary antibody recognizing JNK1 and JNK3 (JNK1/3) (1:200 dilution in blocking solution) for 1 h at room tem-perature. Cells were then rinsed in PBS and incubated with the anti-rabbit secondary antibody for 1 h at room temperature. The 4′,6-diamidino-2-phe-nylindole (DAPI) dye was used for nuclear staining. Cells were rinsed in PBS and mounted, and confocal microscopy was performed.

ChIP assays. ChIP was performed as described before46, starting with 70 µg of chromatin and 5 µg of antibodies recognizing JNK1 and JNK3 (JNK1/3), NF-YA, RNA Pol II, H3K4me2 or H3K27me3. ChIP libraries were prepared with the Illumina ChIP-Seq DNA Sample Prep kit and sequenced on the Genome Analyzer II× following the manufacturer’s protocols. Quantitative PCR after ChIP was performed using SYBR Green chemistry (Applied Biosystems), using 1/40 of the ChIP sample or a 1:100 dilution of input DNA per PCR reaction. Primer sequences are available upon request.

Kinase assays. Recombinant active human JNK1, JNK2 or JNK3 (600 ng; ProQinase) or Aurora B kinase (130 ng; Millipore) were incubated with 100 µM ATP and 0.5 µg of recombinant H3.1 (NEB) in kinase buffer (60 mM HEPES-NaOH, pH 7.5, 5 mM MgCl2, 5 mM MnCl2, 50 µg/ml polyethylene glycol (PEG20,000), 1 mM dithiothreitol (DTT) and 1× PhosSTOP (Roche)). Reactions were incubated at 37 °C for 30 min and terminated by the addition of Laemmli SDS sample buffer. Proteins were separated by 15% SDS-PAGE, and protein blot analysis was performed using rabbit polyclonal antibodies recognizing phosphorylated H3S10 (Millipore) and total H3 (Abcam).

Expression of wild-type and mutant NF-YA in HEK293 cells. Stable cell lines were generated using the pcDNA5/FRT/TO/SH/GW Gateway destina-tion vector and Flp-in HEK293 cells (both from Invitrogen) as described47. ChIP-qPCR was performed after the expression of wild-type or mutant NF-YA for 18 h.

Genomic coordinates. The mouse genome assembly (NCBI37/mm9) was used as a basis for analyses. Annotation of RefSeq transcripts was obtained from the UCSC database (see URLs). Genomic regions were defined as follows: promoter, sequences containing all bases within 1,000 bp of a RefSeq TSS; exon, non-promoter sequences that overlap with exons of RefSeq transcripts; and intron, non-promoter and non-exon sequences flanked by two exons of a single transcript. All other sites were defined as intergenic. A set of non- overlapping TSS regions (N = 17,012, 500 bp upstream and 200 bp downstream of the TSS) was generated using RefSeq TSSs.

Read filtering, alignment and weighting. Low-complexity reads were removed based on dinucleotide entropy (<1% of the reads). Reads were aligned to the mouse genome using Bowtie (version 0.9.9.1)48 with parameters -v 2 -a -m 100, which will find up to 100 best matches for each read with two or fewer mismatches. To track reads without genomic template (for example, exon-exon junctions), reads were also aligned to a databases containing known mouse sequences (see URLs), tracking all best hits with no more than two

mismatches. All quantifications were based on alignments weighted by the inverse of the number of query hits, ensuring that the total weight of a read did not exceed one.

Peak finding. Clusters of ChIP-Seq read alignments were identified using MACS (version 1.3.7.1)49 with a pool of alignments from all biological repli-cates and cellular stages (weights rounded to integers) and parameters set as: mfold = 8, gsize = 2700000000 and tsize = 36. Immunoprecipitation enrich-ment of resulting peak candidates was calculated, and peak candidates with enrichments less than twofold above background were removed, resulting in 902, 480 and 3,335 JNK peaks in ES, NP and TN cells, respectively.

Calculation of peak enrichments in genomic regions. Enrichment of peaks in genomic regions was calculated as the ratio of the observed over the expected number of peaks, where the observed number is the count of all peaks over-lapping a region by more than half of their length and the expected number is the fraction of genomic bases in that region type multiplied by the total number of peaks.

Calculation of immunoprecipitation enrichment. Immunoprecipitation enrichment of a genomic region (TSS window or peak region) was calculated as E = log2 ((nFG / NFG × min(NFG,NBG) + p) / (nBG / NBG × min(NFG,NBG) + p)), where nFG and nBG are the summed weights of overlapping foreground and background (input chromatin) alignments, NFG and NBG are the total number of aligned reads in foreground and background samples, and p is a pseudocount constant (p = 16) used to regularize enrichments with low counts dominated by sampling noise.

Cut-offs for TSS-window enrichments (JNK = 0.7 and NF-YA = 0.5) were manually defined using scatter plots comparing biological replicates to sepa-rate correlated higher enrichments (positives) from the presumably negative and uncorrelated lower enrichments.

RNA-Seq data analysis. RNA from ES, NP and TN cells in two independent biological replicates each was used for cDNA preparation with oligo(dT) prim-ers followed by sequencing on an Illumina Gene Analyzer II×. Expression levels of RefSeq transcripts were calculated by log2 (n / l × avg._l + 1), where n is the weighted sum of alignments to a RefSeq transcript scaled by the total number of reads in the sample, l is the length of the transcript and avg._l is the mean length of RefSeq transcripts. Expressed transcripts were defined based on the bimodal distribution of expression levels as transcripts with expression levels of at least 4.0 (log2-transformed, length-normalized number of reads).

Motif analysis. To identify overrepresented oligonucleotide words, we counted the occurrence of heptamers in the top 1,000 JNK1/3 peaks sorted by immuno-precipitation enrichment in TNs, with TSS windows of expressed genes that did not overlap a JNK peak (N = 7,369) as background. Background counts were scaled to the total number of words in the JNK peaks, and word enrichments was calculated as log2 ((nFG + p) / (nBG + p)), where nFG and nBG are the word counts in the foreground and background sets and p is a pseudocount constant (p = 8). NF-Y–like, SP-1–like and AP-1–like heptamers were identified using the following regular expressions or reverse complements: [AG]CCAAT|CCAAT[CA], CCCCGC|CGCCCC and TGA[CG]TCA, where brackets indicate several alter-native bases that can be found at that position and bars indicate alternative pat-terns that match to forward and reverse-complement sequence words.

Motif weight matrices were identified using MEME version 4.3.0 (ref. 50), with parameters -dna, -mod oops, -w 9 and -revcomp on the top 1,000 JNK1/3 peaks sorted by ChIP enrichment in TNs and a zero-order Markov model background estimated from TSS windows without JNK peaks. Two motifs with E values lower than 1.0 were found, with the GCCCCG (SP-1 like) and CCAAT (NF-Y like) consensus sequences. The AP-1 motif was obtained by running MEME with identical parameters as above on the 18 AP-1 binding sequences downloaded from the JASPAR database51 for motif MA0099.2. Motif hits in TSS windows were obtained using MAST52 with default parameters and the same background as used for MEME.

Prediction of JNK enrichment at TSSs using linear models. Linear models were constructed using R to predict JNK enrichment at TSS windows by linear

© 2

012

Nat

ure

Am

eric

a, In

c. A

ll ri

gh

ts r

eser

ved

.

Nature GeNetics doi:10.1038/ng.1036

combinations of some or all of the variables, including TSS enrichments of RNA Pol II, H3K27me3, H3K4me2 or NF-YA, mRNA expression levels and the number of hits to the CCAAT and GCCCCG motif weight matrices in the TSS window.

Affymetrix microarray data analysis. RNA was processed with the WT cDNA Synthesis and Amplification kit, and labeling was performed with the WT Terminal Labeling kit (both from Affymetrix) according to the manufacturer’s instructions. Labeled RNA was hybridized to GeneChip Mouse Gene 1.0 ST arrays following the protocol in the GeneChip Whole Transcript (WT) Sense Target Labeling Assay manual (Affymetrix) with a hybridization time of 16 h. The Affymetrix Fluidics protocol FS450_0007 was used for washing. Scanning was performed with Affymetrix GCC Scan Control v. 3.0.1 on a GeneChip Scanner 3000 with autoloader (Affymetrix).

Microarrays were RMA-normalized using R Bioconductor53 and the oligo package version 1.14.0 (ref. 54). For each condition, probe set expression levels were averaged across replicate arrays, and fold changes in expression between cells exposed to JNK inhibitor and DMSO were calculated. Probe sets were linked to RefSeq transcripts using Affymetrix annotation (NetAffx release 31, 16 November 2009), retaining a single probe set per gene. Transcripts were split into ten groups of equal size by average expression level in the DMSO-exposed samples, and fold changes in expression of JNK or NF-YA targets within each

expression bin were compared to the non-targets in the same bin using a double-sided Wilcoxon rank-sum test with continuity correction. For visuali-zation, fold changes in all expression bins were median centered at zero.

46. Mohn, F. et al. Lineage-specific polycomb targets and de novo DNA methylation define restriction and potential of neuronal progenitors. Mol. Cell 30, 755–766 (2008).

47. Glatter, T., Wepf, A., Aebersold, R. & Gstaiger, M. An integrated workflow for charting the human interaction proteome: insights into the PP2A system. Mol. Syst. Biol. 5, 237 (2009).

48. Langmead, B., Trapnell, C., Pop, M. & Salzberg, S.L. Ultrafast and memory-efficient alignment of short DNA sequences to the human genome. Genome Biol. 10, R25 (2009).

49. Zhang, Y. et al. Model-based analysis of ChIP-Seq (MACS). Genome Biol. 9, R137 (2008).

50. Bailey, T.L. & Elkan, C. Fitting a mixture model by expectation maximization to discover motifs in biopolymers. Proc. Int. Conf. Intell. Syst. Mol. Biol. 2, 28–36 (1994).

51. Bryne, J.C. et al. JASPAR, the open access database of transcription factor-binding profiles: new content and tools in the 2008 update. Nucleic Acids Res. 36, D102–D106 (2008).

52. Bailey, T.L. & Gribskov, M. Combining evidence using p-values: application to sequence homology searches. Bioinformatics 14, 48–54 (1998).

53. Gentleman, R.C. et al. Bioconductor: open software development for computational biology and bioinformatics. Genome Biol. 5, R80 (2004).

54. Carvalho, B.S. & Irizarry, R.A. A framework for oligonucleotide microarray preprocessing. Bioinformatics 26, 2363–2367 (2010).

© 2

012

Nat

ure

Am

eric

a, In

c. A

ll ri

gh

ts r

eser

ved

.