PNAS PLUS

Mapping of transcription factor motifs in active chromatin identifies IRF5 as key regulator in classical Hodgkin lymphoma Stephan Krehera,b,1, M. Amine Bouhlelc,1, Pierre Cauchyd, Björn Lamprechta,b,e, Shuang Lia,b, Michael Grauf, Franziska Hummela,b, Karl Köcherta,b, Ioannis Anagnostopoulosg, Korinna Jöhrensg, Michael Hummele,g, John Hiscotth, Sören-Sebastian Wenzelb, Peter Lenzf, Markus Schneideri, Ralf Küppersi, Claus Scheidereita, Maciej Giefingj,k, Reiner Siebertj, Klaus Rajewskya, Georg Lenzb, Peter N. Cockerilld, Martin Janza,b, Bernd Dörkena,b,e, Constanze Boniferd,2, and Stephan Mathasa,b,e,2 a Max-Delbrück-Center for Molecular Medicine, 13125 Berlin, Germany; bHematology, Oncology, and Tumor Immunology, Charité–Universitätsmedizin Berlin, 13353 Berlin, Germany; cLeeds Institute of Molecular Medicine, University of Leeds, Leeds LS9 7TF, United Kingdom; dSchool of Cancer Sciences, College of Medical and Dental Sciences, University of Birmingham, Birmingham B15 2TT, United Kingdom; eGerman Cancer Consortium (DKTK), German Cancer Research Center (DKFZ), 69120 Heidelberg, Germany; fDepartment of Physics, Philipps University, 35052 Marburg, Germany; gInstitute of Pathology, Charité– Universitätsmedizin Berlin, 10117 Berlin, Germany; hVaccine and Gene Therapy Institute of Florida, Port St. Lucie, FL 34987; iInstitute of Cell Biology (Cancer Research), University of Duisburg-Essen, 45122 Essen, Germany; jInstitute of Human Genetics, Christian-Albrechts University Kiel and University Hospital Schleswig-Holstein, Campus Kiel, 24105 Kiel, Germany; and kInstitute of Human Genetics, Polish Academy of Sciences, 60-479 Poznan, Poland

Deregulated transcription factor (TF) activities are commonly observed in hematopoietic malignancies. Understanding tumorigenesis therefore requires determining the function and hierarchical role of individual TFs. To identify TFs central to lymphomagenesis, we identified lymphoma type-specific accessible chromatin by global mapping of DNaseI hypersensitive sites and analyzed enriched TFbinding motifs in these regions. Applying this unbiased approach to classical Hodgkin lymphoma (HL), a common B-cell–derived lymphoma with a complex pattern of deregulated TFs, we discovered interferon regulatory factor (IRF) sites among the top enriched motifs. High-level expression of the proinflammatory TF IRF5 was specific to HL cells and crucial for their survival. Furthermore, IRF5 initiated a regulatory cascade in human non-Hodgkin B-cell lines and primary murine B cells by inducing the TF AP-1 and cooperating with NF-κB to activate essential characteristic features of HL. Our strategy efficiently identified a lymphoma type-specific key regulator and uncovered a tumor promoting role of IRF5.

lineages. However, the nature of the TFs initiating and maintaining HRS-specific gene expression remains poorly understood. As an unbiased approach for the identification of deregulated TF activities central to lymphoma biology, we identified HL-specific accessible chromatin regions by global mapping of DNaseI hypersensitive sites (DHSs). DHSs mark cis-regulatory elements bound by TF complexes (9) and differ between normal and malignant cells (10, 11). We then performed an unbiased genomewide search for TF binding motifs enriched within HRS-specific Significance Human lymphomas and leukemias are characterized by molecular and structural alterations of transcription factors (TFs). The identification of such deregulated TFs is therefore central to the understanding of lymphomagenesis. We addressed this question in classical Hodgkin lymphoma (HL), a common B-cell– derived malignancy that is one of the most prominent examples for complex patterns of deregulated TFs including the activation of NF-κB or AP-1 and a profound deregulation of lineagespecific TFs. We found that IRF5 together with NF-κB induces a number of HL characteristic features in non-Hodgkin cells, such as expression of cytokines and chemokines or AP-1 activation. Our work exemplifies how the global lymphoma typespecific characterization of TF activities can improve the understanding of tumor biology.

T

ranscription factor (TF) activities have to be tightly controlled because their aberrant regulation alters tissue-specific gene expression programs and contributes to cancer pathogenesis. Therefore, the identification of altered TF activities in malignancies is of crucial importance to understand malignant transformation and to develop new treatment strategies. Deregulated TF activities are commonly observed in hematopoietic malignancies including human lymphomas and leukemias, and the link between structural or functional alterations in TFs and malignant transformation has been documented in various in vitro and in vivo studies (1–3). Apart from the direct modulation of cellular processes like cellular growth or cell death, alterations of the activity of even single TFs might enforce malignant transformation by switching differentiation programs and consequently altering the cellular fate of the respective cells, as exemplarily demonstrated for the B-lymphoid TF PAX5 (4, 5). Among lymphoid malignancies, one of the most prominent examples for complex patterns of deregulated TFs is classical Hodgkin lymphoma (HL), a common B cell-derived malignancy (6). Pathogenic hallmarks of the malignant Hodgkin/Reed-Sternberg (HRS) cells of HL include the constitutive activation of TFs that are only transiently activated in normal B cells, such as nuclear factor kappa B (NF-κB) or activator protein-1 (AP-1), and a profound deregulation of lineage-specific TFs such as E2A (6–8). Thus, although originating from B-lymphoid cells, HRS cells have lost their B cell-specific gene expression pattern and instead upregulate expression of genes characteristic for other hematopoietic

www.pnas.org/cgi/doi/10.1073/pnas.1406985111

Author contributions: S.K., M.A.B., P.C., S.L., G.L., P.N.C., M.J., C.B., and S.M. designed research; S.K., M.A.B., P.C., B.L., S.L., F.H., S.-S.W., and M.S. performed research; S.K., M.A.B., P.C., B.L., S.L., M. Grau, F.H., K.K., I.A., K.J., M.H., J.H., S.-S.W., P.L., M.S., R.K., C.S., M. Giefing, R.S., K.R., G.L., P.N.C., M.J., B.D., C.B., and S.M. analyzed data; C.B. and S.M. wrote the paper; S.K., M.A.B., P.C., R.K., C.S., M. Giefing, R.S., K.R., G.L., P.N.C., M.J., and B.D. contributed to writing of the manuscript; K.K. analyzed microarray data; I.A. and K.J. performed and interpreted IHC analyses; J.H. provided material; J.H., C.S., R.S., K.R., and B.D. interpreted data; and C.B. and S.M. supervised the project. The authors declare no conflict of interest. This article is a PNAS Direct Submission. Freely available online through the PNAS open access option. Data deposition: The datasets reported in the manuscript have been deposited in the Gene Expression Omnibus (GEO) database, www.ncbi.nlm.nih.gov/geo (accession nos. GSE51726, GSE51813, GSE51717, and GSE51719). 1

S.K. and M.A.B. contributed equally to this work.

2

To whom correspondence may be addressed. Email: [email protected] or stephan. [email protected].

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. 1073/pnas.1406985111/-/DCSupplemental.

PNAS | Published online October 6, 2014 | E4513–E4522

MEDICAL SCIENCES

Edited by Louis M. Staudt, National Cancer Institute, National Institutes of Health, Bethesda, MD, and approved September 4, 2014 (received for review April 17, 2014)

accessible chromatin and functionally analyzed the role of the corresponding factors. The identification of binding motifs for the TFs AP-1, NF-κB, and STAT in HL-specific accessible chromatin confirmed their essential role for HL biology (6). In addition, we revealed that IFN regulatory factor (IRF) binding motifs are among the top enriched motifs in HL. We detected in HRS cells an abundant expression and increased activity of IRF5, which is an IRF TF family member that plays a central role in Toll-like receptor (TLR)-mediated immune responses and is a key regulator of TLR-induced proinflammatory gene expression (12, 13). IRF5 directed HL-specific cytokine expression in cooperation with NFκB and protected HRS cells from cell death. Alone or in combination with NF-κB, IRF5 was capable of inducing gene expression alterations in non-Hodgkin and primary murine splenic B cells that were reminiscent of those found in HL. Furthermore, we identified a transcriptional cross-talk between IRF5 and AP-1, as IRF5 directed the activation of the known abundant and HL-specific AP-1 complex. These data describe a powerful method for the identification of deregulated TF activities in human lymphoma, which led to the identification of IRF5 as a key regulator of HRS cell biology. Results Definition of an HL-Specific Active Chromatin Landscape. To identify transcriptional key regulators of HL biology, we mapped DHSs in both Hodgkin and non-Hodgkin cells, followed by the functional analysis of TFs binding to corresponding TF binding motifs enriched within lymphoma type-specific accessible chromatin (Fig. 1A). Due to the low number of HRS cells in affected lymph nodes (6), we applied this strategy to three classical HLderived cell lines (L1236, L428, and L591; in the following referred to as HRS cell lines) in comparison with two non-Hodgkin B-cell lines (Reh and Namalwa; in the following referred to as non-Hodgkin or NH cell lines; Fig. 1 and SI Appendix, Fig. S1 and Table S1). These cell lines have been successfully used for the identification of key molecular and genomic defects in HL (8, 14, 15). Permeabilized cells were treated with DNaseI, and small genomic fragments were isolated and subjected to next-generation sequencing (DNaseI-Seq) (16). DNaseI-Seq patterns for selected genes are shown in Fig. 1B and SI Appendix, Fig. S1A. We next determined the ratio of tag counts at DHSs between pairs of HRS and NH cell lines and ranked them according to their change in DNaseI-seq signal (Fig. 1C). We concentrated on distal (nonpromoter) sites, as they comprise the majority (about 75–80%) of all DHSs. Many of these elements are enhancers that are more likely to bind tissue-specific factors than promoters (17). The DHS rankings compared with the NH cell lines revealed highly similar DHS accessibility profiles for each of the three HRS cell lines (Fig. 1C). We next divided DHSs into three groups, based on fold change (Fig. 1D and SI Appendix, Fig. S1B): 7,800 NH-specific peaks (marked in green as A); 12,027 shared peaks (marked in black as B); and 6,959 HRS-specific peaks (marked in red as C). To validate the functional relevance of these data with respect to gene expression, we identified the genes closest to the respective DHSs and determined their corresponding mRNA expression level in HRS cell lines by mining microarray expression data (Fig. 1 E and F and SI Appendix, Fig. S1 C–E). In agreement with the requirement of accessible chromatin for ongoing transcription, genes associated with HRSspecific DHSs displayed a higher expression level in HRS than in NH cell lines, irrespective of which HRS was compared with which NH cell line. For both distal and promoter DHSs, the fold increase in HRS-specific nuclease accessibility correlated well with relative gene expression levels (Fig. 1G). Furthermore, promoter DNA methylation levels showed an expected inverse correlation with accessibility at promoter and nearby distal DHSs in L1236 and L428 HRS cell lines (Fig. 1H and SI Appendix, Fig. S1 F and G). Taken together, these data defined a specific active chromatin landscape in HRS cells. E4514 | www.pnas.org/cgi/doi/10.1073/pnas.1406985111

Identification of TF-Binding Motifs Enriched in HRS Cell-Specific Accessible Chromatin Regions. To identify TFs responsible for

the HRS-specific gene expression program, we searched for TF DNA binding motifs specifically enriched in HRS-specific DHSs, using both L1236 and L428 cells as references (Fig. 2 A and B). In the B (NH)-specific DHS signature (Fig. 1D; A), binding motifs for TFs important for the B-cell program such as E-box and OCT motifs (18) were enriched (Fig. 2 A and B), in accordance with their role in maintaining the B-cell phenotype. In contrast, HRS-specific DHSs (Fig. 1D; C) were enriched in binding motifs for members of the inducible TF families AP-1, NF-κB, STAT, and IRF (Fig. 2 A and B), whereas, as previously demonstrated (19), TF motifs important for the B-cell program were found with decreased frequency. Mapping the different binding motifs back to enriched sequences showed their association with NH- or HRS-specific DHSs, respectively, and demonstrated that they clustered around the central position of the DHS (SI Appendix, Fig. S2 A–D). Motif distribution was not random, as AP-1, STAT, and IRF motifs showed a preferred distance with respect to the NF-κB binding motif (SI Appendix, Fig. S2E). A comparison of the DNaseI cutting frequency at the binding motifs for AP-1, NF-κB, and IRF across all HRS- and NH-specific, as well as shared DHSs, revealed that these motifs were protected from DNaseI digestion in HRS cells only, suggesting that they were occupied by a protein complex (Fig. 2 C and D). In contrast, RUNX sites were not protected in any of the analyzed cell types, and CTCF sites were only protected in DHSs shared between HRS and NH cells, in accordance with an invariant binding of CTCF in different blood cell types (20). Abundant IRF5 Expression and Activation in HRS Cells. AP-1, NF-κB, and STAT TFs play an important role in HL pathogenesis (6), and our data confirmed their crucial role in shaping active chromatin in HRS cells. Because IRF binding motifs ranked among the top motifs within HRS-specific DHSs, we investigated IRF expression and function in HL in detail. First, we determined mRNA expression of all known human IRF genes, IRF1–IRF9, in HRS and NH cell lines (Fig. 3A, Upper). IRF1–IRF3 and IRF6–IRF9 were expressed at similar levels in all cell lines, whereas IRF8 expression appeared to be lower in HRS cell lines (Fig. 3A, Upper). IRF4 was robustly expressed in HRS cells as described, but not exclusively (21). Notably, IRF5 was highly expressed in all of the HRS cell lines in accordance with previously published microarray data (22). The IRF5 expression levels exceeded that in non-Hodgkin cell lines including ABC-type diffuse large B-cell lymphoma (DLBCL) cell lines, in which IRF5 mRNA expression has been shown previously (22) (Fig. 3A and SI Appendix, Fig. S3A). We confirmed an activation of the IRF5 locus in HRS cells at the chromatin level (SI Appendix, Fig. S3 B–E), demonstrating an increase in DNaseI accessibility across the entire IRF5 5′-regulatory region (SI Appendix, Fig. S3B), an enrichment of histone H3 lysine 4 trimethylation (H3K4me3) (SI Appendix, Fig. S3C), and the presence of RNA polymerase II (RNA-Pol II) (SI Appendix, Fig. S3D), reflecting ongoing transcription; in contrast, the repressive H3K9me3 mark was absent (SI Appendix, Fig. S3E). In NH cell lines, the IRF5 gene displayed only limited accessibility with low-level enrichment of RNA-Pol II and H3K4me3 (SI Appendix, Fig. S3 B–E), which is consistent with a primed state for an inducible gene. Northern blotting analysis showed abundant and selective expression of two major IRF5 variants (23) in HRS cell lines (SI Appendix, Fig. S3F), corresponding to high IRF5 protein expression (Fig. 3A, Lower). IRF TFs are activated by phosphorylation in the cytoplasm and subsequent translocation into the nucleus (24, 25). In HRS cell lines, we found IRF5 activation as indicated by its nuclear localization (Fig. 3A, Lower) and the presence of DNA-binding IRF5 complexes (Fig. 3B and SI Appendix, Fig. S3 G–I). We confirmed HL-specific IRF5 expression by immunohistochemistry of various primary human lymphomas. IRF5 was abundantly and Kreher et al.

CD79B 75

75

L428

Identify HL-specific DH sites

L591 Namalwa

Determine HL-specific enriched TF binding motifs

Reh

Test biological relevance of corresponding TFs

0 75

L428

0 75

0 75

L591

0 75

0 75

Namalwa

0 75

0 75

Reh

0

0

Mammal conservation

1 0

1 0

L1

L1

23 L4 6 28 L5 91 R eh N am al

23 L4 6 28 L5 91 R eh N am al

w

w

a

Mammal conservation

L1236

0 75

a

L1236 DNaseI HS mapping in HL and non-HL cell lines

PNAS PLUS

CCL5

B

A

Gene expression

Gene expression

200

80

Reh L1236

40

2 0 -2

28 eh /R D eh LM KM - 2 -H /Re L5 2 / R h 40 eh L5 /Re 40 h N Cy / am R al eh w a/ R eh

L4

MEDICAL SCIENCES

0 --200 0

R

eh 40

4

6/

6/ Reh L1236

23

23 L1 80

***

6

+1 0 -1

+4 0 -4

Increasing DNaseI-seq FC

200

2.0

-4

DNaseI-seq FC

mRNA expression FC

***

−4

0

−2

0

2

w n u do p w n u do p w n u do p w n u do p w n u do p w n up

mRNA expression FC (log2)

Distal regions

-1.0

eh 6/ R 28 eh /R N am eh al w a/ R L4

R +4 0 -4

−2.0

DNaseI-seq FC (log2)

-0.5

DNaseI-seq FC L1236/Reh

6/ L1

0.0

1.0 0.5

eh

.

r=0.707

23

2.0

1.5 r=0.732

+0.36 0 -0.36

Increasing DNaseI-seq FC

***

H

Promoters mRNA expression FC (log2)

R C

40 L5

40

y/

eh /R

eh

eh

/R ***

L5

***

L4 28

-H KM

LM D H

2/

-2

R

/R

eh R L1

23

6/

***

1.5

do

mRNA expression FC

***

eh

G eh

F

23

200

Distance to max. DHS (bp)

L1

0 --200 0

do

w

C

0 --200 0

-0.4 0.4

n up

B

40

DHS reads per bp

Increasing DNaseI-seq FC

Increasing DNaseI-seq FC

-4

Reh L1236

80 A

DNaseI tag count FC

+4 0 -4

R

L1 23 G 6/R ro up eh s

+4 0

L1

87 12 M 23

6/

G

eh /R

91 L5

-2.0

E

L1

eh

eh

R

/R

6/

28

23 L1

L4

2.0

D

8

C

-0.4 0.4

H

-2.0

0.0

−1.5 −15

0

15

DNA methylation FC (nearest TSS)

DNaseI-seq FC (log2)

Fig. 1. Definition of an HL-specific active chromatin landscape. (A) Experimental strategy. (B) Screenshots of the CCL5 (Left) and CD79B (Right) loci depicting DNaseI-Seq profiles of HRS (red) and NH (green) cell lines. Underneath, microarray-based gene expression data for the respective genes. (C) HRS cell lines show similar patterns of distal DHSs. The L1236 over Reh DNaseI-Seq fold change (FC) sorted by increasing values and corresponding signals from comparisons L428 or L591 over Reh, respectively, are shown. Spearman rank correlation coefficients compared with L1236/Reh FCs were 0.82 for both L428/Reh and L591/ Reh, respectively. Far right, comparison using the ENCODE lymphoblastoid GM12878 DNaseI-Seq dataset showed a similar result (r = 0.87). (D) Classification of HRS- and NH-specific distal DHSs showing the L1236 over Reh DNaseI-Seq FC (left heat map). Bar next to the heat map: NH (green; group A) and HRS (red; group C) -specific DHSs, respectively. Nonvarying DHSs: black (group B). Panels next to the bar: average DNaseI-Seq profiles for each group. Outermost right, box plot demonstrating the significance of differences in DNaseI-Seq FC for NH- and HRS- specific DHSs (***P < 2.2 × 10−16). (E) The L1236 over Reh increasing DNaseI FC correlates with the FC of mRNA expression of the nearest gene of multiple HRS cell lines compared with the NH cell line Reh (Spearman rank correlation coefficients: 0.73, 0.61, 0.51, 0.5, 0.49, and 0.45, respectively), but there is no such correlation with the NH cell line Namalwa (r = −0.09). (F) Box plots depicting the correlation in mRNA expression FC in HRS cell lines over Reh of nearest genes to sites with low (NH-specific sites, green) and high (HRSspecific sites, red) L1236 over Reh DNaseI-Seq FC. ***P < 2.2 × 10−16. (G) mRNA expression FC (L1236 vs. Reh) correlates positively with DNaseI-seq fold change in both promoter (Upper) and distal (Lower) regulatory regions. Linear fits are indicated by dashed red lines, Spearman correlation coefficients are shown in red. (H) Changes in chromatin accessibility inversely correlate with changes in DNA methylation levels. The heat maps show the L1236 over Reh DNaseI FC sorted by increasing L1236 over Reh DNaseI ratio and the methylation FC of the corresponding element, respectively. Spearman rank correlation coefficients compared with L1236/Reh DHS FC were −0.45, −0.303, and 0.041 for L1236/Reh, L428/Reh, and Namalwa/Reh m5C FC, respectively.

consistently detectable in the nucleus and cytoplasm of HRS cells (37 of 38 HL cases) and was absent in the vast majority of non-Hodgkin lymphomas (Fig. 3C and SI Appendix, Fig. S3J and Kreher et al.

Table S2). Although IRF5 expression was observed in several DLBCL cases, nuclear IRF5 staining in DLBCL was only occasionally detected (5 of 45 cases) without preference for ABC- or PNAS | Published online October 6, 2014 | E4515

B

DNaseI-seq FC L1236/Reh

A

DNaseI-seq FC L428/Reh

down

up

down

up

ETS, p=10-103, 21.57% of seq.

AP-1, p=10-458, 25.00% of seq.

ETS, p=10-171, 16.68% of seq.

AP-1, p=10-761, 48.58% of seq.

E-Box, p=10-103, 35.73% of seq.

STAT, p=10-204, 29.93% of seq.

OCT, p=10-123, 5.25% of seq.

RUNX, p=10-206, 35.33% of seq.

OCT, p=10-97, 2.41% of seq.

NF-κB, p=10-182, 4.92% of seq.

E-Box, p=10-115, 26.38% of seq.

NF-κB, p=10-139, 18.26% of seq.

PU.1 + IRF, p=10-48, 2.41% of seq.

IRF, p=10-182, 4.31% of seq.

PU.1 + IRF, p=10-78, 3.54% of seq.

STAT, p=10-36, 15.97% of seq.

RUNX, p=10-46, 7.87% of seq.

BLIMP1, p=10-91, 12.21% of seq.

RUNX, p=10-27, 6.10% of seq.

IRF, p=10-36, 3.28% of seq.

DNaseI-seq FC L1236/Reh

C

invariant

down

DNaseI-seq FC L428/Namalwa

D up

down L1236

0.01

invariant

up L428

0.006

Reh

Namalwa

IRF 0 -100

0

100

-100

0

100

-100

0

IRF 0 -100

100

0.005

0

100

-100

0

100

-100

0

100

0.003

0

100

-100

0

100

-100

0

100

0.01

AP-1 0 -100

0

100

-100

0

100

-100

0

100

NF-κB DNaseI cut density

DNaseI cut density

NF-κB 0 -100

0 -100

0

100

-100

0

100

-100

0

100

0.01

AP-1 0 -100

0

100

-100

0

100

-100

0

100

0.01

0.01

RUNX 0 -100

0

100

-100

0

100

-100

0

RUNX 0 -100 0.01

100

0

100

-100

0

100

-100

0

100

0.012

CTCF 0 -100

0

100

-100

0

100

-100

0

100

Distance to motif centre (bp)

CTCF 0 -100

0

100

-100

0

100

-100

0

100

Distance to motif centre (bp)

Fig. 2. Enriched TF binding motifs in HRS cell-specific accessible chromatin regions. (A and B) HRS-specific DHSs contain motifs for inducible TFs. Sequences within HRS (red) and NH (green) -specific DHSs (A, L1236 over Reh; B, L428 over Reh) as defined in Fig.1D were analyzed for the presence of enriched TF binding motifs. (C and D) Digital footprinting using average DNaseI cutting frequencies ± 100 bp around the HRS-enriched binding motifs IRF, NF-κB, and AP-1 for (C) L1236 vs. Reh and (D) L428 vs. Namalwa (red and green, respectively) in HRS-specific, shared (invariant), and NH-specific DHSs, corresponding to groups A, B, and C. Note the differences in average DNaseI cutting frequency in DHSs specific for HRS and NH cell lines and the reduction in cutting frequency around the motif centers. RUNX and CTCF motifs were examined as controls demonstrating (i) equal digestion around RUNX motifs in DHS of all cell types and (ii) increased digestion around CTCF motifs only within shared DHS.

GCB-type DLBCL. Furthermore, IRF5 expression in HRS cells was stronger than that observed in tonsils of patients with EBVcaused infectious mononucleosis (SI Appendix, Fig. S3J). IRF5 Controls Cytokine Activation in HRS Cells and Protects Them from Cell Death. IRF5 attracted our attention because it orches-

trates the transcriptional activation of proinflammatory genes in response to various stimuli (12, 26), and a broad activation of proinflammatory response genes is a unique hallmark of HL (6). To define the role of IRF5 in establishing the proinflammatory phenotype of HRS cells, we first determined the mRNA expression levels of a panel of cytokine, chemokine, and chemokine receptor genes in the various cell lines. We included in this E4516 | www.pnas.org/cgi/doi/10.1073/pnas.1406985111

analysis the TFs JUN, a known regulator of cytokine expression, and LPS-induced TNF-α factor (LITAF), which mediates LPSinduced cytokine and chemokine activation, and the LPS/IRFinducible chemokine gene CXCL11, the latter two genes not yet analyzed in HRS cells (Fig. 3D). All these genes were consistently expressed in HRS but not in NH cell lines. IRFs and NF-κB can form a multiprotein complex that is required for full target gene activation (13, 27). Given that persistent activation of NF-κB is a key pathogenic feature of HRS cells (6), we speculated that their unusual cytokine and chemokine production resulted from IRF5 activation in combination with NF-κB. To explore this possibility, we analyzed in HRS cell lines the activities of wild-type (WT) or mutated reporter Kreher et al.

L4

PNAS PLUS L5

N

IR F IC 5

IR F IC 5

IR F IC 5

-D H

40

L1 7

SU

6 23

L4 L428

RT-PCR

IR F IC 5

n.s. free probe

C

HL

WC

IRF5

N

40 ** * 20

pG L3 IS RE wt N F mu t -κ Bm ut

** ** **

1 0

RANTES

1.2 L540Cy 0.8

******

** 0.4

*** *** ***

******

+

+

+ ++

+

+

+ ++

***

+

+ + ++

G Percentage of viable GFP+ cells

IL6

** * ** *

L428

6

** * ** ** **

2 0

RANTES P

L540Cy

80

80 ** ** * *

60 40

* 60 ** ** * 40

20

20

0 hours after cell sorting

Mock IκBαΔN ** *

RANTES P

RANTES P

IL13

KM-H2

10 **

3 ** *

14

** ** * ** *

** ** **

siCTL siIRF5_#1 siIRF5_#2

0 0

72

96

0

48

72

Fig. 3. Abundant IRF5 expression in HRS cells regulates cytokine expression and protects from cell death. (A, Upper) Analysis of IRF1 to IRF9 and, as a control, GAPDH mRNA expression by RT-PCR. (Lower) IRF5 protein expression in whole cell (WC) and nuclear (N) extracts. β-actin and PARP were analyzed as controls. One of four independent experiments is shown. (B) EMSA of complexes bound to the ISRE without (−) or with addition of IRF5 antibody or its isotype control (IC) [supershift (ss)]. n.s., nonspecific complex. Underneath, free probe of the gel. One of four independent experiments is shown. (C) Representative IHC analysis of IRF5 in classical HL. HRS cells showing nuclear and cytoplasmic IRF5 staining are marked by arrows. (D) RT-PCR analysis of various genes, as indicated. GAPDH was analyzed as control. One of four independent experiments is shown. (E) IRF and NF-κB binding sites are required for full RANTES promoter (P) activation. pGL3-basic, pGL3-RANTES P WT, or pGL3-RANTES P constructs with mutated ISRE or NF-κB site were transfected into HRS cell lines, together with Mock plasmid (filled bars) or IκBαΔN (open bars). Luciferase activity is shown as fold activation compared with pGL3-basic (set 1). Data are represented as mean ± SD. One of six independent experiments is shown. (F) IRF5 and NF-κB are required for endogenous IL13, IL6, and RANTES mRNA expression. L540Cy cells were transfected with control plasmid (Mock) or DNIRF5-4D– and/or IκBαΔN-expression plasmids. mRNA expression in enriched transfected cells was analyzed in comparison with the Mock control, set as 1, by real-time PCR. Error bars denote 95% CIs. One of six independent experiments is shown. (G) L428 and L540Cy cells were transfected with control (siCTL) or IRF5-specific siRNA (siIRF5_#1 and siIRF5_#2) constructs, and the percentage of viable cells was determined at the indicated times in enriched transfected cells. Data are represented as mean ± SD. One of four independent experiments is shown. P values are shown for the comparisons to the respective WT reporter activites or the Mock controls, respectively. *P < 0.05; **P < 0.01; ***P < 0.001.

constructs of the RANTES promoter (RANTES P), which contains an IFN-stimulated response element (ISRE) and an NF-κB binding site (27) (Fig. 3E). In accordance with the high RANTES mRNA expression in these cells (Fig. 3D), the WT construct was highly active (Fig. 3E). ISRE mutation reduced promoter activity, and mutation of the NF-κB site reduced activity to basal levels. Furthermore, cotransfection with the NF-κB superrepressor IκBαΔN (28) reduced activities of the WT- and ISRE-mutated RANTES P constructs (Fig. 3E), suggesting a cooperative activity of IRFs and NF-κB in HRS cell lines. To determine the impact of IRF5 and NF-κB on the endogenous expression of proinflammatory genes in Kreher et al.

HRS cells, we inhibited both factors by cotransfecting the dominant-negative variants DNIRF5-4D (24) and IκBαΔN, respectively (Fig. 3F and SI Appendix, Fig. S4 A and B). Transfection of DNIRF5-4D and IκBαΔN independently reduced IL13, IL6, and RANTES mRNA expression, and both in combination resulted in a more pronounced reduction in mRNA expression (Fig. 3F and SI Appendix, Fig. S4A). These effects were confirmed by analyzing IL-6 and RANTES protein expression in supernatants of transfected HRS cell lines (SI Appendix, Fig. S4C). In addition to controlling the expression of these proinflammatory genes, IRF5 was essential for HRS cell viability. Following transient IRF5 knockdown PNAS | Published online October 6, 2014 | E4517

MEDICAL SCIENCES

*

5

0

F

0 Mock DNIRF5-4D IκBαΔN

L1236

L3 w IS RE t N F mu t -κ Bm ut

Luciferase activiy (fold activation)

60

pG

cell line IL6 IL13 TNFA LTA CSF2 RANTES CXCL10 CXCL11 CCR7 LITAF JUN GAPDH

E

w IS RE t NF mu t -κ Bm ut

L4 L128 2 KM 36 L5 -H 91 2 H D L5 LM L 5 40 2 R 40C e Nh y a BL ma l BJ -60 wa SUAB -D H L4

Hodgkin cell lines

pG L3

PARP

cytokine relative to GAPDH mRNA expression

ss ss IRF5

IRF5 ß-actin

D

cell line

L4 L1 2 8 KM236 L 5 -H 2 H 91 D L5 L M L5 4 0 -2 R 40C e Nh y a BL ma l B J - 60 w a SUAB -D H L4

cell line IRF1 IRF2 IRF3 IRF4 IRF5 IRF6 IRF7 IRF8 IRF9 GAPDH

IR F IC 5

Hodgkin cell lines

am al w

a

B 28

A

a subset of proinflammatory genes in both cell lines, full activation of most genes to HRS-cell–like expression levels required their combined activity. We demonstrated this synergistic induction exemplarily for RANTES and IL-6 also at the protein level (Fig. 4B and SI Appendix, Fig. S5B). Functionally, the production of cytokines and chemokines following transfection of NH cell lines (Fig. 4A) led to an increased attraction of mononuclear cells (Fig. 4C), in accordance with the chemoattraction of various normal hematopoietic cells by HRS cells (6). Finally, we performed microarray profiling of IRF5-4D–, IKKβ(EE)-, and IRF5-4D– together with IKKβ(EE)-transfected Reh cells and subsequent gene set enrichment analysis (GSEA). Evaluation of HRS signatures based on genes specifically up-regulated in HRS cell lines (Fig. 4D and SI Appendix, Table S3) and primary HRS cells (SI Appendix, Fig. S5C and Table S3) revealed a significant enrichment of the HRS signature in IRF5-4D or IKKβ(EE) single transfectants, with the highest enrichment following transfection of both factors in combination (Fig. 4D and SI Appendix, Fig. S5C).

using specific small-interfering (si)RNAs, we observed an induction of cell death of ∼35–50% of the transfected HRS cell population (Fig. 3G and SI Appendix, Fig. S4 D and E). The analysis of the cells by Annexin-V/propidium iodide (PI) and active caspase-3 staining indicated that cell death occurred by induction of apoptosis (SI Appendix, Fig. S4 F and G). IRF5 Together with NF-κB Activates an HRS Cell-Like Inflammatory Gene Expression Pattern in Non-Hodgkin Cells. Our results promp-

ted the question of whether IRF5 alone or in combination with NF-κB was sufficient to induce the HL characteristic proinflammatory gene expression program. In a subsequent series of experiments, we therefore mimicked the HRS-cell–characteristic activities of IRF5 and NF-κB by introducing the constitutively active variants IRF5-4D (29) and IKKβ(EE) (30), respectively, into non-Hodgkin cells. In HEK293 cells, either IRF5 or NF-κB activation activated the RANTES P construct, but both together induced a dramatic synergistic activation (SI Appendix, Fig. S5A). We confirmed the synergistic activation of endogenous proinflammatory gene expression in the IRF5-4D- and/or IKKβ(EE)transfected NH cell lines Reh and BJAB (Fig. 4 and SI Appendix, Fig. S5 B and C). Although IRF5-4D or IKKβ(EE) alone induced

B

Reh

IL-6 (pg/ml)

cytokines

L1

M 236 o IR ck

IL13 TNFA LTA

Reh

Reh ** *

60

F IK 5-4 K D IR β(E IK F5- E) K 4 L1 β(ED+ 2 E M 36 ) o IR ck F IK 5-4 K D IR β(E IK F5- E) Kβ 4 D (E + E)

BJAB

cell line IL6

to initiate HRS-cell–characteristic gene expression in primary B

40

RANTES (pg/ml)

A

IRF5 Induces Key Aspects of the HL-Specific Gene Expression Program in Primary Murine Splenic B Cells. To assess the potential of IRF5

**

20

1200

** *

800 400

ns * 0 Mock + + + IRF5-4D ++ IKKβ(EE)

0 Mock + + + IRF5-4D ++ IKKβ(EE)

chemokines/ chemokine-r.

CSF2 CXCL10 CXCL11 CCR7 LITAF

C

TFs

JUN GAPDH

D

BJAB

Reh

multifold spontaneous migration

RANTES

**

4 3

4

** ** **

3 ns

2

2

ns

1

1

0 Mock + + + IRF5-4D ++ IKKβ(EE)

0

+

+

+ ++

Mock Zero cross at 10554

-4

IRF5 0

NES -2.742 FDR, 0.000 -0.6

10,000 20,000 Rank in ordered dataset

IRF5-4D

2

Mock

0 -4

0.0

NES -2.936 FDR, 0.000

-0.7

Zero cross at 11410

IKK 0

10,000 20,000 Rank in ordered dataset

IKKβ(EE)

List ranked metric (signal to noise)

2 0

0.0

Enrichement score (ES)

Enrichement score (ES)

NES -2.033 -0.40 FDR, 0.000

List ranked metric (signal to noise)

List ranked metric (signal to noise)

Enrichement score (ES)

HRS-specific up-regulated genes 0.10 0.00

2.5

Mock (positively correlated)

0.0 -5.0 0

Mock

Zero cross at 11290

IRF5/IKK 20,000 10,000 Rank in ordered dataset

IRF5-4D+IKKβ(EE)

Enrichment profile Hits Ranking metric scores

Fig. 4. IRF5 together with NF-κB determines the inflammatory phenotype of HRS cells. (A) IRF5 alone or with NF-κB activates HL-specific inflammatory genes in NH cells. BJAB and Reh cells were transiently transfected with control plasmids (Mock) or IRF5-4D– and/or IKKβ(EE)-expression plasmids. Enriched transfected cells were analyzed for mRNA expression of the indicated genes by RT-PCR. L1236 cells and GAPDH expression served as controls. One of three independent experiments is shown. (B) IL-6 and RANTES secretion by Mock and IRF5-4D and/or IKKβ(EE) transfected Reh cells analyzed by ELISA. One of four independent experiments is shown. (C) Chemotaxis assay of peripheral blood mononuclear cells toward supernatants (SN) of IRF5-4D and/or IKKβ(EE) transfected Reh and BJAB cells. Fold migration is shown as relating to spontaneous migration toward medium of Mock-transfectants, set 1.0. One of three independent experiments is shown. (D) GSEA enrichment plots of the HRS cell line signature (SI Appendix, Table S3; signature based on the comparison of HRS vs. NH cell lines, log2 FC cutoff 2) of up-regulated genes of IRF5-4D–, IKKβ(EE)-, or IRF5-4D–, and IKKβ(EE)-induced gene expression changes in transiently transfected Reh cells. Note, that the absolute value of NES is highest in IRF5-4D plus IKKβ(EE) cells, indicative of an additive or synergistic effect of the two factors with respect to a Hodgkin-like phenotype shift of transfected Reh cells. Error bars denote SDs. P values are shown for comparisons to the respective Mock controls. ns, not significant; *P < 0.05; **P < 0.01; ***P < 0.001.

E4518 | www.pnas.org/cgi/doi/10.1073/pnas.1406985111

Kreher et al.

HRS-specific down-regulated genes 0.5

HRS-specific up-regulated genes Enrichement score (ES)

Enrichement score (ES)

A

affected B cell-associated genes and TFs, as demonstrated for down-regulation of AICDA and the TFs EBF1, PRDM1, and XBP1, as well as their antagonists, as shown for the upregulation of ABF1 and ID2 (7) in a HRS-cell–characteristic manner (Fig. 5B). Other typical features of HRS cells were also recapitulated by IRF5 activity such as up-regulation of CD30 (also called TNFRSF8) or the c-MET tyrosine-kinase receptor and silencing of the epigenetic regulator CBFA2T3 (31) (Fig. 5B).

NES 1.927 FDR 0.000 0.0

PNAS PLUS

cells, we infected purified murine splenic B cells with control (Mock) or IRF5-4D encoding retroviruses (Fig. 5 and SI Appendix, Fig. S5 D and E). We characterized IRF5-4D– induced gene expression changes by microarray analyses and evaluated the HRS signature by GSEA (Fig. 5A and SI Appendix, Fig. S5D). This analysis revealed a significant enrichment of the up- and down-regulated genes of the HRS signature in IRF5-infected B cells. Apart from the induction of proinflammatory genes, IRF5-induced gene deregulation

Enrichment profile Hits Ranking metric scores

0.1 0.0 NES -1.827 FDR 0.000

IRF5 10,000 20,000 Rank in ordered dataset

0

2.5 0.0

Mock (positively correlated) Mock Zero cross at 10276

IRF5 10,000 20,000 Rank in ordered dataset

-5.0 0

IRF5-4D

IRF5-4D

LTA CCL5 CXCL10 CXCL11 CCR7

F5

IR k F5 -4 D cytokines/ chemokines and receptors

IL6 TNFA

cell line

L4 2 L1 8 2 KM 36 R -H2 eh N a M mal oc w IR k a

#2

splenic NH B cells

HL

M oc

#1

M o

B

IR ck F5 -4 D

C

-4 D

-5.0

Zero cross at 10276

AP-1 n.s. Sp1

n.s.

LITAF IRF5-4D–transduced splenic B cells

D

IRF4 PRDM1 XBP1 CD30

n.s.

E

#1

M

CBFA2T3 GAPDH

cJU JU N pa NB nJU N

L4 2 R 8 eh AP-1

#2

IR k F5 -4 D

BCL6

4D

PAX5 AICDA

ss ss

IR k F5 -

E2A

oc

ID2 EBF1

B cell transcription factors and regulators

ABF1

M oc

c-MET

JUN JUNB ATF3

Fig. 5. IRF5 orchestrates HRS-specific gene expression in primary murine B cells and acts upstream of AP-1. (A) GSEA enrichment plots of IRF5-4D–induced gene expression changes in primary murine B cells of the HRS cell line signature for down (Left) and up (Right) -regulated genes are shown. Note, that IRF5-4D induces a highly significant and specific shift of murine splenic B cells to a Hodgkin-like phenotype. (B) IRF5-4D–induced mRNA expression changes of various HRS-characteristic genes were analyzed by RT-PCR. GAPDH was analyzed as control. Two (#1 and #2) of four independent experiments are shown. (C–E) IRF5 up-regulates AP-1 activity in murine splenic B cells. (C) Whole cell extracts of Mock- and IRF5-4D–transduced B cells were analyzed by EMSA for AP-1 DNA binding activity. L428, L1236, and KM-H2 HRS and Reh and Namalwa NH cells, and TF Sp1 DNA binding activity were included as controls. Note that the nonspecific complex is detectable in human but not murine cells and that AP-1 activation among the cell lines is restricted to HRS ones. One of three independent experiments is shown. n.s., nonspecific complex. (D) IRF5-4D–transduced splenic B cells were analyzed by EMSA for AP-1 DNA binding activity without (−) or with addition of antibodies specific for c-JUN, JUNB, or c-JUN, JUNB, and JUND (pan-JUN). L428 and Reh cells were included as controls. One of three independent experiments is shown. ss, supershift. n.s., nonspecific complex. (E) IRF5-4D–transduced splenic B cells were analyzed for JUN, JUNB, and ATF3 mRNA expression by RT-PCR. GAPDH is shown in B. Two (#1 and #2) of three independent experiments are shown.

Kreher et al.

PNAS | Published online October 6, 2014 | E4519

MEDICAL SCIENCES

Mock

2.5 0.0

List ranked metric (signal to noise)

List ranked metric (signal to noise)

-0.5

We next focused on the analysis of the TF AP-1, whose deregulation is a molecular hallmark of HRS cells (28). In agreement with the up-regulation of JUN in NH cell lines (Fig. 4A), IRF5-4D caused a strong induction of AP-1 DNA binding activity (Fig. 5C). The binding intensity and aberrant migration pattern of the IRF5-induced AP-1 complex closely resembled that observed in HRS cells (28), as demonstrated by supershift analyses with JUN antibodies (Fig. 5D). In line with these observations, IRF5-4D induced the AP-1/CREB factor genes JUN, JUNB, and ATF3, which encode three main components of the HL-specific AP-1 complex (28, 32), at the transcriptional level (Fig. 5E), indicating that AP-1 is a downstream effector of IRF5. Discussion The genome-wide mapping of DHSs has been used successfully to characterize cell type-specific regulatory elements, as well as occupied TF binding motifs within these regions (33, 34). Only recently, the comparison of DHSs in various normal and malignant cell types revealed characteristic reprogramming patterns of the epigenome in distinct cancer entities including hematopoietic malignancies (10, 11). Here, we extended such analyses by combining the definition of cell type-specific accessible chromatin regions with the functional characterization of TFs binding to enriched binding motifs in such regions. In the HRS-specific DHSs, we identified enriched binding motifs particularly for inducible TFs including NF-κB, STAT, AP-1, and IRFs. For NF-κB, STAT, and AP-1 factors, an important role not only for lymphomagenesis in general but also for the pathogenesis of HL has been already established (6). For example, NF-κB and STATs are required for growth and survival of the HRS cells, and various genomic defects in respective pathway components confirm their essential role for HL pathogenesis (6). Our data confirm their important role for HRS-specific gene regulation and, vice versa, validated our experimental approach. Given the established importance of these factors for HL biology, we focused in this work on IRF factors, which have important functions in the differentiation and transformation of hematopoietic cells (13). HL represents a lymphoid malignancy with highly complex TF alterations and dramatic changes of the lymphoid-specific gene expression program. The most important result from this study is therefore that the deregulated activity of only a few TFs might be sufficient to initiate this program. The constitutive activity of IRF5 protects HRS cells from cell death and in combination with NF-κB, IRF5 induces key aspects of the HRS-cell–characteristic gene expression program. Even in primary splenic B cells, IRF5 activation results in the abundant induction of proinflammatory genes (6), the down-regulation of genes required for initiation and maintenance of the B-lineage differentiation program including terminal plasma cell differentiation, and the upregulation of their transcriptional antagonists. Moreover, IRF5 promotes the HL-characteristic silencing of the epigenetic regulator CBFA2T3, which is involved in the aberrant activation of DNA long terminal repeat regions in HRS cells (8). We also identified AP-1 as a downstream effector of IRF5. AP-1 deregulation has been recognized as a hallmark of HRS cells, being involved in the HRS cell-specific deregulation of CD30 (35), the immunomodulatory gene LGALS1 (also called galectin-1) (36), and their unique dedifferentiation process (7, 8). In HRS cells AP-1 is activated by an unusual mechanism, as its activation is partly MAPK independent (28, 37), and it is primarily composed of JUN and ATF subunits instead of JUN and FOS (28, 32). The IRF5-mediated transcriptional activation of the AP-1 complex described here positions IRF5 upstream of AP-1 in the transcriptional hierarchy in HRS cells and provides a molecular explanation for the thus far poorly understood AP-1 activation in HL. Whether the IRF5-mediated gene expression alterations, including the induction of the various cytokines and chemokines, modify the constitutive E4520 | www.pnas.org/cgi/doi/10.1073/pnas.1406985111

NF-κB activity in HRS cells, which is in part mediated by secreted factors (38), remains to be investigated. The comparison of the DHSs data of HL and NH cells, as well as an Epstein Barr virus (EBV)-transformed lymphoblastoid cell line dataset from the Encyclopedia of DNA Elements (ENCODE), highlighted an HRS-specific active chromatin landscape. This finding indicates that, although EBV infection is linked to HL pathogenesis (6, 39), it is per se not sufficient to reprogram the epigenome of B-lymphoid cells toward a HRS cell-like conformation. Furthermore, the HL-specific epigenetic landscape significantly correlated with the gene expression pattern in the various HRS cell lines, including those of T-cell origin (HDLM-2 and L540). This result suggests that a common transformation process with dominant transcriptional alterations leads to the HL phenotype rather than that the HL gene expression pattern reflects that of a putative physiological counterpart. The important role of IRF5 in the orchestration of the HL-specific gene expression program points toward potential initiating events in this disease. In innate immunity, IRF5 is positioned downstream of pattern recognition receptors (PRR) (13) and is required for the induction of various inflammatory mediators. For example, IRF5−/− mice are resistant to LPS-induced lethal shock, emphasizing the role of IRF5 in the coordination of innate immune responses (12, 24, 26, 40). In addition, IRF5 gene polymorphisms and expression alterations have been linked to autoimmune and inflammatory diseases such as rheumatoid arthritis and inflammatory bowel diseases (12, 41, 42). With respect to IRF5 activation, an involvement of MyD88-dependent signaling complexes following TLR activation and a role of the IKK-related kinases TBK1 and IKKe have been described (12, 43). Thus, in HRS cells, high-level IRF5 expression might be a result of constitutive PRR activation, a hypothesis that can now be experimentally tested. Full target gene induction by IRF5 requires the cooperation with other TFs or coactivators (13, 44, 45). In line with these data, we observed a strong transcriptional synergism of IRF5 with NF-κB, which only in this combination resulted in full activation of a whole set of proinflammatory genes in NH cells. Our data indicate that this synergism is the basis for the abundant production of cytokines and chemokines in HRS cells, as NF-κB activation alone can be observed in a variety of other hematopoietic malignancies, which, however, lack the widespread and abundant activation of proinflammatory genes characteristic for HL. It remains to be investigated how IRF5 and NF-κB coordinate the protection of HRS cells from apoptotic cell death, and whether other transcription factors aberrantly activated in HL, such as GATA3 (7, 46), AP-1, or the here described LITAF, contribute to full induction of proinflammatory genes in HL. Together, our studies uncovered a key role of IRF5 in a lymphoid malignancy with a unique inflammatory phenotype and exemplifies how the global lymphoma type-specific identification of aberrant TF activities can improve the understanding of tumor biology. Methods Cell Lines, Culture Conditions, and Transfections. HRS [L428, L1236, KM-H2, L591 (EBV+), HDLM-2, L540, L540Cy], pro-B lymphoblastic leukemia (Reh), Burkitt´s lymphoma (Namalwa, BL-60, BJAB), diffuse large B-cell lymphoma (OCI-Ly3, OCI-Ly10, HBL1, TMD8, HT, OCI-Ly1, OCI-Ly7, OCI-Ly19, SU-DHL-4), and HEK293 cell lines were cultured as previously described (8). Cells were electroporated (EP) in OPTI-MEM I using Gene-Pulser II (Bio-Rad) with 950 μF and 0.18 kV (L428, L1236, L540Cy, BJAB, HEK293), 50 μF and 0.5 kV (KM-H2), and 500 μF and 0.3 kV (Reh, L591). Transfection efficiency was determined by pEGFP-N3 (Clontech Laboratories) cotransfection and FACS analysis. Reh and BJAB cells were transfected with 60–70 μg of a pcDNA3-FLAGIRF5-4D and/or 20–40 μg of a pRK5-IKKβ(EE) expression plasmid along with 10 μg pEGFP-N3. L428, L591, and L540Cy cells were transfected with 60 μg of a pcDNA3-FLAG-DNIRF5-4D and/or 40 μg of a pcDNA3-IκBαΔN expression plasmid or controls or with 50 μg of pSUPER plasmid (47)-based siIRF5

Kreher et al.

DNA Constructs. Expression constructs pcDNA3-IκBαΔN and pRK5-IKKβ(EE) and luciferase reporter constructs pGL3-RANTES P WT, pGL3-RANTES P-ISRE mut, and -NF-κB mut have been previously described (8, 27, 28). To generate the pcDNA3-FLAG-IRF5-4D expression construct, full-length human IRF5 cDNA (GenBank accession no. U51127) was cloned with N-terminal FLAG epitope into pcDNA3 (Invitrogen). Activating mutations according to Lin et al. (29) were introduced by site-directed mutagenesis using the QuikChange Multi Site-Directed Mutagenesis Kit (Stratagene). pcDNA3-FLAG-DNIRF54D lacks AA 1–137, which comprise the N-terminally located nuclear localization signal and the DNA binding domain and was cloned through BamHI and EcoRI restriction sites into pcDNA3-FLAG from pcDNA3-FLAGIRF5-4D. For cloning of the retroviral expression construct MSCV-IRF54D-IRES-CFP, the plasmid pcDNA3.1/Zeo-IRF5-4D (29) was digested with BamHI followed by a Klenow fill-in reaction and EcoRI cleavage. The resulting IRF5-4D fragment was cloned into the vector MSCV-IRES-CFP (a gift of F. Rosenbauer, Institute of Molecular Tumor Biology, University of Münster, Germany) via EcoRI and XhoI (the latter converted to a blunt end by Klenow fill-in). Sequences for IRF5 siRNAs were selected based on previously described design rules (48). siRNA constructs were generated by cloning target sequences siRNA_2113 5′-GAGTGAGCACTTAGGTATCAT-3′ (siIRF5 #1) and siRNA_1025 5′-GTGGAACTCTTCGGCCCCATA-3′ (siIRF5 #2) through BglII and HindIII restriction sites into pSUPER (47). The scrambled siRNA construct has been previously described (8). All constructs were verified by sequencing.

1. Rui L, Schmitz R, Ceribelli M, Staudt LM (2011) Malignant pirates of the immune system. Nat Immunol 12(10):933–940. 2. Rosenbauer F, Tenen DG (2007) Transcription factors in myeloid development: Balancing differentiation with transformation. Nat Rev Immunol 7(2):105–117. 3. Mullighan CG (2013) Genome sequencing of lymphoid malignancies. Blood 122(24): 3899–3907. 4. Cobaleda C, Jochum W, Busslinger M (2007) Conversion of mature B cells into T cells by dedifferentiation to uncommitted progenitors. Nature 449(7161):473–477. 5. Cobaleda C, Busslinger M (2008) Developmental plasticity of lymphocytes. Curr Opin Immunol 20(2):139–148. 6. Küppers R (2009) The biology of Hodgkin’s lymphoma. Nat Rev Cancer 9(1):15–27. 7. Mathas S, et al. (2006) Intrinsic inhibition of transcription factor E2A by HLH proteins ABF-1 and Id2 mediates reprogramming of neoplastic B cells in Hodgkin lymphoma. Nat Immunol 7(2):207–215. 8. Lamprecht B, et al. (2010) Derepression of an endogenous long terminal repeat activates the CSF1R proto-oncogene in human lymphoma. Nat Med 16(5):571–579. 9. Cockerill PN (2011) Structure and function of active chromatin and DNase I hypersensitive sites. FEBS J 278(13):2182–2210. 10. Ptasinska A, et al. (2012) Depletion of RUNX1/ETO in t(8;21) AML cells leads to genome-wide changes in chromatin structure and transcription factor binding. Leukemia 26(8):1829–1841. 11. Stergachis AB, et al. (2013) Developmental fate and cellular maturity encoded in human regulatory DNA landscapes. Cell 154(4):888–903. 12. Takaoka A, et al. (2005) Integral role of IRF-5 in the gene induction programme activated by Toll-like receptors. Nature 434(7030):243–249. 13. Tamura T, Yanai H, Savitsky D, Taniguchi T (2008) The IRF family transcription factors in immunity and oncogenesis. Annu Rev Immunol 26:535–584.

Kreher et al.

DNaseI Digestion. To perform DNaseI accessibility assays with HRS and NH cell lines, the optimal cell number and DNaseI (DPFF DNaseI; Worthington Biochemical Corporation) concentration were titrated for each cell line. The DNA digestion extent was comparable in all of the generated samples as measured by RT-PCR (10). Briefly, 1 × 106 to 5 × 106 cells were permeabilized with detergent and immediately digested at 3 × 107 cells per mL for 3 min at 22 °C with DNaseI at 1–12 U/mL in digestion buffer supplemented with 1 mM CaCl2 (10). The nuclei were lysed, the nuclear proteins were digested with 1 mg/mL Proteinase K overnight at 37 °C, and the DNA was isolated by phenol/chloroform extraction. Levels of DNaseI digestion were assessed using real-time PCR, measuring the ratio of presence of known DNaseI hypersensitive regions to more resistant gene-free regions. Sequences of realtime PCR primers used were, for the active region, TBP promoter 5′CTGGCGGAAGTGACATTATCAA and 5′-GCCAGCGGAAGCGAAGTTA, and for the inactive region, a gene free control region of chromosome 18: 5′ACTCCCCTTTCATGCTTCTG and 5′-AGGTCCCAGGACATATCCATT. DNaseI-Seq samples were generated from a size selection of DNaseI-digested DNA fragments comprised within a range of 100–600 bp and subjected to library preparation as per the manufacturer´s instructions (Illumina). Libraries were run on an Illumina GAIIx sequencer. More detailed information is provided in SI Appendix, SI Methods. ACKNOWLEDGMENTS. We thank S. Kressmann and B. Wollert-Wulf for outstanding technical assistance, P. Rahn for cell sorting, F. Rosenbauer for providing the MSCV-IRES-CFP construct, and S. A. Assi and D. Westhead for hosting the DNaseI data and providing data links. This work was supported in part by Deutsche Forschungsgemeinschaft Grants SFB/TRR54 and JA1847/1-1 (to S.M. and M.J.) and KU1315/7-1 and GRK1431 (to R.K.), the Berliner Krebsgesellschaft, the Experimental and Clinical Research Center, a joint cooperation between the Charité-Universitätsmedizin Berlin and the Max Delbrück Center for Molecular Medicine, and the German Cancer Consortium. Research in Leeds and Birmingham was supported by grants from Cancer Research UK (to C.B.), as well as Leukemia Lymphoma Research (to C.B. and P.N.C.). M. Giefing was supported by an Federation of European Biochemical Societies Long-Term Fellowship and Support for International Mobility of Scientists fellowship of the Polish Ministry of Sciences and Higher Education.

14. Steidl C, et al. (2011) MHC class II transactivator CIITA is a recurrent gene fusion partner in lymphoid cancers. Nature 471(7338):377–381. 15. Rui L, et al. (2010) Cooperative epigenetic modulation by cancer amplicon genes. Cancer Cell 18(6):590–605. 16. Leddin M, et al. (2011) Two distinct auto-regulatory loops operate at the PU.1 locus in B cells and myeloid cells. Blood 117(10):2827–2838. 17. Heinz S, et al. (2010) Simple combinations of lineage-determining transcription factors prime cis-regulatory elements required for macrophage and B cell identities. Mol Cell 38(4):576–589. 18. Matthias P, Rolink AG (2005) Transcriptional networks in developing and mature B cells. Nat Rev Immunol 5(6):497–508. 19. Overbeck BM, et al. (2012) ETS1 encoding a transcription factor involved in B-cell differentiation is recurrently deleted and down-regulated in classical Hodgkin’s lymphoma. Haematologica 97(10):1612–1614. 20. Lichtinger M, et al. (2012) RUNX1 reshapes the epigenetic landscape at the onset of haematopoiesis. EMBO J 31(22):4318–4333. 21. Falini B, et al. (2000) A monoclonal antibody (MUM1p) detects expression of the MUM1/IRF4 protein in a subset of germinal center B cells, plasma cells, and activated T cells. Blood 95(6):2084–2092. 22. Küppers R, et al. (2003) Identification of Hodgkin and Reed-Sternberg cell-specific genes by gene expression profiling. J Clin Invest 111(4):529–537. 23. Barnes BJ, Moore PA, Pitha PM (2001) Virus-specific activation of a novel interferon regulatory factor, IRF-5, results in the induction of distinct interferon alpha genes. J Biol Chem 276(26):23382–23390. 24. Barnes BJ, Kellum MJ, Field AE, Pitha PM (2002) Multiple regulatory domains of IRF-5 control activation, cellular localization, and induction of chemokines that mediate recruitment of T lymphocytes. Mol Cell Biol 22(16):5721–5740.

PNAS | Published online October 6, 2014 | E4521

PNAS PLUS

Motif Discovery, Heatmaps, and Distributions. We used the findMotifsGenome function in Homer for primary motif detection, where de novo discovery was performed in regions ranging [−200bp; +200bp] around the maximum DHS. Motif reinjection was carried out in RSA Tools Matrix-Scan (49). Resulting outputs were converted to gff format and subsequently computed as frequencies every 10 bp for all regions and ordered according to fold changes. Heatmaps were generated via Java Treeview. Motif densities were computed relative to each DHS maximum, where distances used were that between the start of each motif (regardless of the strand) and the DHS maximum. Motif distances to NF-κB were obtained using BedTools closest and plotted in R.

MEDICAL SCIENCES

expression plasmids or respective scrambled siRNA controls along with 10 μg pEGFP-N3 for enrichment of transfected cells by FACS sorting, or additionally along with 10 μg of an expression plasmid encoding truncated mouse MHC class I molecule H-2Kk (Miltenyi Biotec) for enrichment by MACS sorting. Forty-eight to 72 h after transfection, GFP+ cells were enriched by FACS sorting or MACS sorting using of MACSelect Kk MicroBeads according to the manufacturer´s recommendations (Miltenyi Biotech). For analysis of luciferase activity, L428, L1236, KM-H2, and HEK293 cells were transfected by EP with 10 μg of reporter constructs, together with 100 ng pRL-TKLuc as an internal control. Where indicated, cells were additionally transfected with pcDNA3-IκBαΔN, pRK5-IKKβ(EE), and/or pcDNA3-FLAG-IRF5-4D expression constructs or the respective controls. Twenty-four to 48 h after transfection, the ratio of the two luciferases was determined (Dual luciferase kit; Promega). Murine splenic B cells were isolated from C57BL6/J mice, purified by CD43 depletion (Miltenyi Biotec), and cultured in the presence of 20 μg/mL LPS (Sigma), all according to standard protocols. After 24 h, cells were retrovirally transduced with MSCV-IRF5-4D-IRES-CFP (IRF5-4D) expression plasmid or MSCV-IRES-CFP (Mock) as a control. Forty-eight hours after retroviral transduction, CFPpositive cells were FACS sorted, and mRNA or protein extracts were prepared. The use of human material was approved by the Ethikkommision of the Charité - Universitätsmedizin Berlin and was performed in accordance with the Declaration of Helsinki. Animal care and experiments were performed in accordance with the institutional guidelines.

25. Chen W, et al. (2008) Insights into interferon regulatory factor activation from the crystal structure of dimeric IRF5. Nat Struct Mol Biol 15(11):1213–1220. 26. Barnes BJ, et al. (2004) Global and distinct targets of IRF-5 and IRF-7 during innate response to viral infection. J Biol Chem 279(43):45194–45207. 27. Génin P, Algarté M, Roof P, Lin R, Hiscott J (2000) Regulation of RANTES chemokine gene expression requires cooperativity between NF-kappa B and IFN-regulatory factor transcription factors. J Immunol 164(10):5352–5361. 28. Mathas S, et al. (2002) Aberrantly expressed c-Jun and JunB are a hallmark of Hodgkin lymphoma cells, stimulate proliferation and synergize with NF-kappa B. EMBO J 21(15):4104–4113. 29. Lin R, Yang L, Arguello M, Penafuerte C, Hiscott J (2005) A CRM1-dependent nuclear export pathway is involved in the regulation of IRF-5 subcellular localization. J Biol Chem 280(4):3088–3095. 30. Delhase M, Hayakawa M, Chen Y, Karin M (1999) Positive and negative regulation of IkappaB kinase activity through IKKbeta subunit phosphorylation. Science 284(5412): 309–313. 31. Lamprecht B, Bonifer C, Mathas S (2010) Repeat-element driven activation of protooncogenes in human malignancies. Cell Cycle 9(21):4276–4281. 32. Janz M, et al. (2006) Classical Hodgkin lymphoma is characterized by high constitutive expression of activating transcription factor 3 (ATF3), which promotes viability of Hodgkin/Reed-Sternberg cells. Blood 107(6):2536–2539. 33. Boyle AP, et al. (2008) High-resolution mapping and characterization of open chromatin across the genome. Cell 132(2):311–322. 34. Neph S, et al. (2012) An expansive human regulatory lexicon encoded in transcription factor footprints. Nature 489(7414):83–90. 35. Watanabe M, et al. (2003) AP-1 mediated relief of repressive activity of the CD30 promoter microsatellite in Hodgkin and Reed-Sternberg cells. Am J Pathol 163(2): 633–641. 36. Juszczynski P, et al. (2007) The AP1-dependent secretion of galectin-1 by Reed Sternberg cells fosters immune privilege in classical Hodgkin lymphoma. Proc Natl Acad Sci USA 104(32):13134–13139.

E4522 | www.pnas.org/cgi/doi/10.1073/pnas.1406985111

37. Habelhah H, et al. (2004) Ubiquitination and translocation of TRAF2 is required for activation of JNK but not of p38 or NF-kappaB. EMBO J 23(2):322–332. 38. Krappmann D, et al. (1999) Molecular mechanisms of constitutive NF-kappaB/Rel activation in Hodgkin/Reed-Sternberg cells. Oncogene 18(4):943–953. 39. Hjalgrim H, et al. (2003) Characteristics of Hodgkin’s lymphoma after infectious mononucleosis. N Engl J Med 349(14):1324–1332. 40. Krausgruber T, et al. (2010) IRF5 is required for late-phase TNF secretion by human dendritic cells. Blood 115(22):4421–4430. 41. Graham RR, et al.; Argentine and Spanish Collaborative Groups (2006) A common haplotype of interferon regulatory factor 5 (IRF5) regulates splicing and expression and is associated with increased risk of systemic lupus erythematosus. Nat Genet 38(5):550–555. 42. Krausgruber T, et al. (2011) IRF5 promotes inflammatory macrophage polarization and TH1-TH17 responses. Nat Immunol 12(3):231–238. 43. Schoenemeyer A, et al. (2005) The interferon regulatory factor, IRF5, is a central mediator of toll-like receptor 7 signaling. J Biol Chem 280(17):17005–17012. 44. Lenardo MJ, Fan CM, Maniatis T, Baltimore D (1989) The involvement of NF-kappa B in beta-interferon gene regulation reveals its role as widely inducible mediator of signal transduction. Cell 57(2):287–294. 45. Taniguchi T, Takaoka A (2002) The interferon-alpha/beta system in antiviral responses: A multimodal machinery of gene regulation by the IRF family of transcription factors. Curr Opin Immunol 14(1):111–116. 46. Stanelle J, Döring C, Hansmann ML, Küppers R (2010) Mechanisms of aberrant GATA3 expression in classical Hodgkin lymphoma and its consequences for the cytokine profile of Hodgkin and Reed/Sternberg cells. Blood 116(20):4202–4211. 47. Brummelkamp TR, Bernards R, Agami R (2002) A system for stable expression of short interfering RNAs in mammalian cells. Science 296(5567):550–553. 48. Reynolds A, et al. (2004) Rational siRNA design for RNA interference. Nat Biotechnol 22(3):326–330. 49. Turatsinze JV, Thomas-Chollier M, Defrance M, van Helden J (2008) Using RSAT to scan genome sequences for transcription factor binding sites and cis-regulatory modules. Nat Protoc 3(10):1578–1588.

Kreher et al.