Gene regulation and suppression of type I interferonsignaling by STAT3 in diffuse large B cell lymphomaLi Lua,b,1, Fen Zhua,b,1, Meili Zhangc,d,1, Yangguang Lia,b, Amanda C. Drennana,b, Shuichi Kimparaa,b, Ian Rumballa,b,Christopher Selzera,b, Hunter Camerona,b, Ashley Kellicuta,b, Amanda Kelma,b, Fangyu Wanga,b,Thomas A. Waldmannc,2, and Lixin Ruia,b,2

aDepartment of Medicine, University of Wisconsin School of Medicine and Public Health, Madison, WI 53792; bCarbone Cancer Center, University ofWisconsin School of Medicine and Public Health, Madison, WI 53792; cLymphoid Malignancies Branch, Center for Cancer Research, National Cancer Institute,National Institutes of Health, Bethesda, MD 20892; and dLaboratory Animal Science Program, Leidos Biomedical Research, Inc., Frederick, MD 21702

Contributed by Thomas A. Waldmann, December 7, 2017 (sent for review August 28, 2017; reviewed by Vu N. Ngo and Demin Wang)

STAT3 is constitutively activated in many cancers and regulatesgene expression to promote cancer cell survival, proliferation,invasion, and migration. In diffuse large B cell lymphoma (DLBCL),activation of STAT3 and its kinase JAK1 is caused by autocrineproduction of IL-6 and IL-10 in the activated B cell–like subtype(ABC). However, the gene regulatory mechanisms underlying thepathogenesis of this aggressive lymphoma by STAT3 are not wellcharacterized. Here we performed genome-wide analysis andidentified 2,251 STAT3 direct target genes, which involve B cellactivation, survival, proliferation, differentiation, and migration.Whole-transcriptome profiling revealed that STAT3 acts as botha transcriptional activator and a suppressor, with a comparablenumber of up- and down-regulated genes. STAT3 regulates multi-ple oncogenic signaling pathways, including NF-κB, a cell-cyclecheckpoint, PI3K/AKT/mTORC1, and STAT3 itself. In addition,STAT3 negatively regulates the lethal type I IFN signaling pathwayby inhibiting expression of IRF7, IRF9, STAT1, and STAT2. Inhibitionof STAT3 activity by ruxolitinib synergizes with the type I IFN in-ducer lenalidomide in growth inhibition of ABC DLBCL cells in vitroand in a xenograft mouse model. Therefore, this study provides amechanistic rationale for clinical trials to evaluate ruxolitinib or aspecific JAK1 inhibitor combined with lenalidomide in ABC DLBCL.

STAT3 | interferon | diffuse large B cell lymphoma

Diffuse large B cell lymphoma (DLBCL), the most commonnon-Hodgkin lymphoma, includes two main molecular sub-

types: an activated B cell–like (ABC) and a germinal center B cell–like (GCB) (1, 2). ABC DLBCL is more aggressive, with autocrinesignaling from the cytokines IL-6 and IL-10 that constitutively ac-tivates JAK1 (3) and STAT3 (4–7) to promote cell survival. ABCDLBCL cells have high NF-κB activity (8) due to genetic alterationsin the Toll-like receptor (TLR) and B cell receptor signalingpathways (9–11). Somatic mutations of MYD88 (mainly L265P), akey signaling adaptor in TLR signaling, engage the NF-κB pathwayand induce production of IL-6, IL-10, and IFNβ (11). In contrast toIL-6 and IL-10, IFNβ is proapoptotic, and its basal level is low toundetectable in ABC DLBCL cells (12). IFNβ production can beprevented by the transcription factors IRF4 and SPIB through re-pression of IRF7, a transcription factor for IFNβ expression (12).Recently, we revealed that, in addition to STAT3 activation,

JAK1 is present in the nucleus and directly phosphorylates his-tone H3 on tyrosine 41 (H3Y41) to induce expression of nearly3,000 genes, including the NF-κB pathway genes MYD88 andIRF4 (3, 13). These findings suggest a positive feedback loopbetween the cytokine and NF-κB signaling pathways that pro-motes the malignant phenotype of ABC DLBCL cells. Thisepigenetic gene regulatory mechanism by JAK1 is distinct fromthe canonical JAK-STAT signaling pathway, given that morethan 90% of these genes do not bear a STAT motif in theirpromoter region (3). Using STAT3 chromatin immunoprecipi-tation followed by DNA sequencing (ChIP-seq) and whole-genome transcriptome (RNA-seq) analysis in ABC DLBCL

cell lines and control GCB DLBCL cell lines that lack STAT3activation, a recent study has identified a total of 10,337 STAT3-binding regions corresponding to 8,531 genes, of which 1,545genes are differently expressed between the two subtypes andlargely associated with ABC DLBCL biology (14). However,biological significance and functional association of these STAT3targets in ABC DLBCL cells remain to be studied.Here, we use genetic and pharmacological inhibition of STAT3 in

ABC DLBCL cells and identify genes that are directly regulated bySTAT3. These STAT3 target genes are involved in multiple sig-naling pathways, which promote cancer cell survival and pro-liferation. In addition, STAT3 negatively regulates lethal type Isignaling by inhibiting expression of IRF7, IRF9, STAT1, andSTAT2, key transcription factors for IFNβ production and signaling.

ResultsGenome-Wide Analysis Identifies STAT3 Transcriptional Target Genesin ABC DLBCL Cells. To identify STAT3 target genes genome-wide,we performed STAT3 ChIP-seq in the ABC DLBCL cell linesTMD8 and OCI-Ly10, in which high levels of STAT3 phosphor-ylation were detected by immunoblot analysis (Fig. 1A). SinceSTAT3 activity was sufficiently inhibited by the JAK1/JAK2 in-hibitor AZD1480 (Fig. 1A) and this inhibitor was used for H3Y41-P ChIP-seq analysis (3), AZD1480-treated cells served as a controlfor STAT3 ChIP-seq experiments. Using the model-based analysis

Significance

We demonstrate that STAT3 is a critical transcriptional regulatorof the activated B cell–like subtype of diffuse large B cell lym-phoma (ABC DLBCL), the most common, aggressive, non-Hodgkinlymphoma. By genome-wide assessment, we have identified tar-get genes of STAT3. Gene regulation by STAT3 in ABC DLBCL ac-centuates survival signaling pathways while dampening the lethaltype I interferon pathway. Knowledge of these STAT3-regulatedgenes has led to our demonstration that a small-molecule in-hibitor in the JAK1-STAT3 signaling pathway synergizes with thetype I interferon inducer lenalidomide, suggesting a new thera-peutic strategy for ABC DLBCL, a subtype that is particularly dif-ficult to treat and has poor prognosis.

Author contributions: L.R. designed research; L.L., F.Z., M.Z., Y.L., A.C.D., S.K., I.R., C.S.,H.C., A. Kellicut, A. Kelm, and F.W. performed research; T.A.W. and L.R. supervised re-search; L.L., F.Z., M.Z., Y.L., S.K., and L.R. analyzed data; and L.L., F.Z., M.Z., T.A.W., andL.R. wrote the paper.

Reviewers: V.N.N., City of Hope National Medical Center; and D.W., BloodCenterof Wisconsin.

The authors declare no conflict of interest.

Published under the PNAS license.1L.L., F.Z., and M.Z. contributed equally to this work.2To whom correspondence may be addressed. Email: tawald@mail.nih.gov or lrui@medicine.wisc.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1715118115/-/DCSupplemental.

E498–E505 | PNAS | Published online January 2, 2018 www.pnas.org/cgi/doi/10.1073/pnas.1715118115

Dow

nloa

ded

by g

uest

on

Mar

ch 2

8, 2

021

of ChIP-seq (MACS) for peak calling (16), we identified a total of11,487 STAT3-binding sites (peaks) in TMD8 cells and 22,856 in

OCI-Ly10 cells compared with the AZD1480-treated control sam-ple (Fig. 1B and Dataset S1). Specificity of these STAT3-binding

A

TMD80 15' 30' 1h 2h 4h 6h 0

OCI-Ly1015' 30' 1h 2h 4h 6h

pSTAT3

STAT3

-actin

AZD1480 (2 M):

DMSO AZD1480TMD8 OCI-Ly10

-5K Apex +5K-5K Apex +5K

DMSO AZD1480

Den

sity

hea

tmap

of 1

1,48

7 pe

aks

Den

sity

hea

tmap

of 2

2,85

6 pe

aks

High

Low

High

Low

STAT3 motif enrichment (OCI-Ly10)STAT3 motif enrichment (TMD8)STAT3 motif enrichment (TMD8)

E-value 3.6e-21 E-value 9.5e-8E

Position of Best Site in SequencePosition of Best Site in Sequence

Pro

babi

lity

Pro

babi

lity

Pro

babi

lity

pSTAT3

STAT3

-actin

AZD1480 (2 M):

B

C

0

DTMD8(4,746)

OCI-Ly10 (6,058)

2,495 3,8072,251

E

0 1 2 3 4

Regulation of metabolic processResponse to external stimulus

Cellular response to stressCell cycle

Regulation of cell migrationApoptotic process

Intracellular signal transductionLymphocyte differentiation

Lymphocyte proliferationB cell activation 7.03E-05

3.87E-03

1.27E-02

5.81E-12

8.49E-04

4.81E-02

5.00E-03

5.29E-04

3.97E-035.38E-06

Enrichment Score

bits

0

1

2

0

1

2

bits

1 2 3 4 5 6 7 8 9 10 11 1 2 3 4 5 6 7 8 9 10 11

Fig. 1. Genome-wide analysis reveals STAT3 transcriptional targets in ABC DLBCL. (A) Immunoblot analysis of phospho-STAT3 and total STAT3 protein inTMD8 and OCI-Ly10 cells treated with 2 μMAZD1480. (B) Heat maps of STAT3 ChIP-seq in TMD8 and OCI-Ly10 cells after 4 h of treatment with either DMSO or2 μM AZD1480. STAT3 peak summits were centered within 5 kb of the flanking sequence on either side. Darker color indicates a higher density of reads.STAT3 peaks were ranked by signal intensity at the peak center, and the same order is used to display the AZD1480-treated sample. (C) The CentriMo plotsshow the distribution of known the STAT3 motif in the ChIP-seq peak summit regions (P < 0.001). De novo motif discovery from STAT3 ChIP-seq peaks showsidentical sequence logs to the known STAT consensus motif (15). (D) Venn diagram shows 2,251 STAT3-binding genes that are shared in TMD8 and OCI-Ly10 cells. (E) Gene ontology analysis of the 2,251 STAT3-binding genes (P < 0.05).

Lu et al. PNAS | Published online January 2, 2018 | E499

MED

ICALSC

IENCE

SPN

ASPL

US

Dow

nloa

ded

by g

uest

on

Mar

ch 2

8, 2

021

sites was confirmed by the MEME motif enrichment analysis (15),with a similar distribution pattern of STAT3 motifs between the twocell lines (Fig. 1C).

Based on genomic loci of these peaks, we mapped near a protein-coding gene within a window extending from −15 kb 5′ of thetranscriptional start site to the 3′ end of any annotated transcript

A TMD8(4,746)

1420

1075

Ctrl #10 #16

2157

1989

OCI-Ly10 (6,058)shSTAT3

Ctrl #10 #16shSTAT3

1

-1

0.5

0

-0.5

shSTAT3 Ctrl #10 #16

IL6-

JAK

-ST

AT

3 an

d N

F-

B g

ene

sign

atur

es

TMD8shRNA shSTAT3

Ctrl #10 #16

OCI-Ly10shRNA

DMSO30

130

130

130

1

AZD1480

DMSO

AZD1480

TM

D8

OC

I-Ly1

0

NFKB2 0 +2K +4K +6K

DMSO30

130

130

130

1

AZD1480

DMSO

AZD1480

TM

D8

OC

I-Ly1

0

STAT3 +20K+40K 0K06+

B

C

TNFAIP8

TRAF1

CD44PTPN1

NFKB2

CXCL13

IL2RA

STAT3

NAMPT

BCL3

CD69

DRAM1

SPSB1

SGK1GPR183FOS

DMSO20

120

120

120

1

AZD1480

DMSO

AZD1480

TM

D8

OC

I-Ly1

0

SOCS3 0+1K+2K+3K

DMSO15

115

115

115

1

AZD1480

DMSO

AZD1480

TM

D8

OC

I-Ly1

0

IL10 0+1K+2K+3K+4K+5K

0.0

0.5

1.0

1.5

******

0.0

0.5

1.0

1.5

****

0.0

0.5

1.0

1.5

*****

0.0

0.5

1.0

1.5

***

Control shRNA

#10 #16 shSTAT3

Control shRNA

#10 #16 shSTAT3

Control shRNA

#10 #16 shSTAT3

Control shRNA

#10 #16 shSTAT3

Rel

ativ

e S

OC

S3

mR

NA

Rel

ativ

e IL

10 m

RN

A

TMD8

TMD8

OCI-Ly10

OCI-Ly10

D DMSO

20

120

120

120

1

AZD1480

DMSO

AZD1480

TM

D8

OC

I-Ly1

0

CD69 +10K 0

Fig. 2. Whole-transcriptome profiling reveals signaling pathways regulated by STAT3 in ABC DLBCL. (A) Heatmaps show expression changes of STAT3-binding genes inTMD8 and OCI-Ly10 cells after 2 d of knockdown of STAT3 by two shRNAs. (B) Heat maps show down-regulation of IL6-JAK-STAT3 and NF-κB pathway genes by twoSTAT3 shRNAs. STAT3-binding genes are shown in red. (C) STAT3 is recruited to regulatory regions of indicated genes, as shown by read density tracks (DMSO controls inred, AZD1480-treated samples in green). (D) Quantitative PCR analysis of SOCS3 and IL-10 mRNA levels (normalized to GAPDHmRNA levels) in TMD8 and OCI-LY10 cellsafter 1 d (only for SOCS3 in TMD8) or 2 d of knockdown of STAT3 by shRNAs. Error bars represent mean ± SD of triplicates (*P < 0.05, **P < 0.01, ***P < 0.001).

E500 | www.pnas.org/cgi/doi/10.1073/pnas.1715118115 Lu et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 2

8, 2

021

associated with the gene, as for our previous study (3). We identi-fied 4,746 potential STAT3 target genes in TMD8 cells and 6,058 inOCI-Ly10 cells, with an overlap of 2,251 genes between the two celllines (Fig. 1D and Dataset S1). Considering these overlapped genesas common STAT3 targets in ABC DLBCL, we performedPANTHER gene ontology analysis (17). The results revealed thatthese common target genes were enriched for biological pro-cesses that include B cell activation, proliferation, differentiation,

cell-cycle progression, stress response, cell migration, and metab-olism (Fig. 1E), suggesting an important role for STAT3 in thepathogenesis of ABC DLBCL.

Whole-Transcriptome Profiling Reveals That STAT3 Acts as both aTranscriptional Activator and a Repressor in ABC DLBCL. To determinegenes that are directly regulated by STAT3, we performed thewhole-transcriptome analysis by RNA-seq in the above TMD8 and

STAT3

pSTAT3

STAT1

pSTAT1

IRF7

IRF9

-actin

TMD8 OCI-Ly10 SUDHL7

A

B

pSTAT3

pSTAT1

STAT1

IRF7

IRF9-actin

shSTAT3

STAT3-C (OCI-Ly10) STAT3-C (HBL1)

Dox

shSTAT3 Ctrl #10 #16

Type 1 IFN pathway

TMD8shRNA shSTAT3

Ctrl #10 #16

OCI-Ly10shRNA

DMSO30

330

330

330

3

AZD1480

DMSO

AZD1480

DMSO18

318

318

318

3

AZD1480

DMSO

AZD1480

STAT1

+10K 0 -10K

STAT2

0+2K+4K

TM

D8

OC

I-Ly1

0T

MD

8O

CI-L

y10

C

pSTAT2

STAT2

1

-1

0.5

0

-0.5

D

_ + _ + _ +

15

115

115

115

1

1

IRF7

15

11115

115

1151515

1

0+1K+2K+3K

IRF9

K4+K2+

1

0

CASP8

IRF1SP110

IRF9ADARCD47PARP14STAT2

IRF2

IL15

STAT1DDX60 RTP4 USP18GBP4

IFI44EIF2AK2EPSTI1OASLNUB1

MX1

TDRD7TRAFD1

IFITM2CD74

CNP CXCL11 CXCL10

40

40

15

1

15

1

1

40

1

15

1

15

1

40

1

40

1

Fig. 3. STAT3 suppresses type I IFN signaling in ABC DLBCL. (A) STAT3 is recruited to regulatory regions of STAT1, STAT2, IRF7, and IRF9, as shown by read densitytracks (DMSO controls in red, AZD1480-treated samples in green). (B) Heat maps show expression changes of type I IFN pathway genes in TMD8 and OCI-Ly10 cellsafter 2 d of knockdown of STAT3 by two shRNAs. STAT3-binding genes are shown in red. (C) Immunoblot analysis of the indicated proteins in TMD8, OCI-Ly10,and the control cell line SUDHL7 after 2 d of knockdown of STAT3 by shRNA. (D) Immunoblot analysis of the indicated proteins in TMD8 and HBL1 cells after 2 or4 d of retroviral expression of constitutively activated STAT3 (STAT3-C).

Lu et al. PNAS | Published online January 2, 2018 | E501

MED

ICALSC

IENCE

SPN

ASPL

US

Dow

nloa

ded

by g

uest

on

Mar

ch 2

8, 2

021

OCI-Ly10 cell lines. We knocked down STAT3 by two shRNAsfrom our previous study (7). As shown in Fig. 2A, 53% (2,495/4,746) of the STAT3 target genes in TMD8 cells and 68% (4,146/

6,058) in OCI-Ly10 cells changed their expression when STAT3was knocked down. Of note, the number of down-regulated geneswas comparable to that of up-regulated genes, suggesting that

TMD8

05

10152025303540

0

5

10

15

20

25

30

0

5

10

15

20

25

30

35

40

45

* ****

*

**

* ***

*

Live

cel

ls (

x 1,

000)

OCI-Ly10 SUDHL7

Day 3Day 0

A

Day 3Day 0Day 0 Day 3

0

0.2

0.4

0.6

0.8

1

1.2

0 2 4 6 8

....................................................................

4 M Ruxolitinib

8 M Ruxolitinib

4 M Ruxolitinib

8 M Ruxolitinib

Lenalidomide ( M)10

Com

bina

tion

inde

x (C

I)

Synergism: CI<1D

E

TMD8

pSTAT3

-actin

Ruxolitinib ( M): 0 1 2 4 0 1 2 4

OCI-Ly10

0

5

10

15

20

25

Contro

l

shSTAT3

0

1

2

3

***

*

Lena

shSTAT3+

L

ena

******

***

****

Rel

ativ

e IF

NB

1 m

RN

A

TMD8 OCI-Ly10

Contro

l

shSTAT3

Lena

shSTAT3+

Lena

B

C

Control

0

0.2

0.4

0.6

0.8

1

1.2

0 2 4 6 8

....................................................................

4 M Ruxolitinib

8 M Ruxolitinib

4 M Ruxolitinib

8 M Ruxolitinib

Lenalidomide ( M)10

Com

bina

tion

inde

x (C

I)

Synergism: CI<1

TMD8 OCI-Ly10

F

3 72 3 76 3 2

DMSO ctrl 4 M Lena 4 M Ruxolitinib4 M Lena plus4 M Ruxolitinib

S:34

G1:46G2/M:9

34 34 29

478 48 8

557

15

72

15 13 14

79

7-AAD

Brd

U

TMD8

OCI-Ly10

shSTAT3LenashSTAT3+Lena

Vehicle control LenalidomideRuxolitinib Combination

Tum

or v

olum

e (m

m )3

Days after initiation of therapy

***

Vehicle control

Lenalidomide

Ruxolitinib

Combination

* ****

*** ***

Fig. 4. Synergism between STAT3 inhibition and lenalidomide in growth inhibition of ABC DLBCL. (A) STAT3 shRNA enhanced lenalidomide-mediatedtoxicity in TMD8 and OCI-Ly10 cells but not in control SUDHL7 cells. The 4,000 cells per well from each cell line were plated on a 96-well plate and inducedwith doxycycline for expression of STAT3 shRNA. No doxycycline cultures served as a control. Total live cells were counted with trypan blue exclusion assayafter 3 d of shRNA induction in the presence of 4 μM lenalidomide or DMSO. Data are presented as mean ± SD of triplicates (*P < 0.05, **P < 0.01).(B) Quantitative PCR analysis of IFNβmRNA levels (normalized to GAPDHmRNA levels) in TMD8 and OCI-Ly10 cells after 3 d of knockdown of STAT3 by shRNAsin the presence of 4 μM lenalidomide or DMSO. Error bars represent mean ± SD of triplicates (*P < 0.05, **P < 0.01, ***P < 0.001). (C) Immunoblot analysis ofphospho-STAT3 and β-actin loading control in TMD8 and OCI-Ly10 cells after 30 min of treatment with the indicated concentrations of ruxolitinib.(D) Synergism between ruxolitinib and lenalidomide in cell killing in TMD8 and OCI-Ly10 cells. Cells were treated with ruxolitinib and lenalidomide for 5 dbefore trypan blue dye exclusion viability assay. Combination index (CI) was calculated with CompuSyn software. (E) Cell-cycle analysis of TMD8 and OCI-Ly10 cells after 5 d of ruxolitinib and lenalidomide. Cells were pulsed with 10 μM BrdU for 3 h before flow cytometric analysis. (F) The combination therapy oflenalidomide with ruxolitinib significantly inhibits OCI-Ly10 tumor growth in vivo. Female NOD/SCID mice were injected s.c. with 1 × 107 OCI-Ly10 cells. Tendays later, lenalidomide was given i.p. at a dose of 15 mg/kg/d for 14 d, and ruxolitinib was continuously administered by an s.c. miniosmotic pump at a doseof 60 mg/kg/d for 14 d. Shown is a photograph of tumors from each group at day 15 (Top) and average tumor volumes during the therapeutic course for eachgroup (Bottom). Error bars show the SD of eight mice/group (*P < 0.05, **P < 0.01, lenalidomide or ruxolitinib group compared with vehicle control group;***P < 0.001, combination group compared with lenalidomide or ruxolitinib group).

E502 | www.pnas.org/cgi/doi/10.1073/pnas.1715118115 Lu et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 2

8, 2

021

STAT3 functions as both a transcriptional activator and a re-pressor in ABC DLBCL cells.Next, we performed gene set enrichment analysis (GSEA) to

identify the signaling pathways in which these up-regulated anddown-regulated genes are involved. The results revealed signif-icant enrichment in gene signature of multiple signaling path-ways (Fig. S1), including the PI3K/AKT/mTORC1 (Fig. S2),E2F/G2M cell-cycle checkpoint (Fig. S3), IL6-JAK-STAT3, NF-κB, and type I IFN signaling pathways. Consistent with previousstudies (4, 5, 14), STAT3 is involved in the positive feedbackregulation of the IL-6/IL-10 signaling and shows crosstalk withNF-κB signaling pathways in ABC DLBCL cells (Fig. 2 B–D).We verified SOCS3, a negative regulator of JAK-STAT signaling(18), as a target gene of STAT3 (Fig. 2D). More significantly,16 of the genes down-regulated by STAT3 shRNAs are directSTAT3 targets, including TNFAIP8, TRAF1, CD44, CD69,BCL3, FOS, SGK1, NF-κB2, and STAT3 itself (Fig. 2 B and C).

STAT3 Suppresses Type I IFN Signaling in ABC DLBCL. In ABC DLBCL,production of the proapoptotic cytokine IFNβ can result fromoncogenic MYD88 mutations (11). This type I IFN signaling issuppressed by the transcription factors IRF4 and SPIB, whichrepress IRF7 expression to prevent IFNβ transcription and TYK2activation (12). Our recent work demonstrated that IRF4 andSPIB are epigenetic targets of JAK1 due to H3Y41 phosphory-lation, but IRF4 expression is not regulated by STAT3 (3). Thesefindings along with the above RNA-seq analysis prompted us toinvestigate whether STAT3 directly targets critical genes in theIFNβ signaling pathway. It is known that, in response to IFNβ,STAT1 and STAT2 are phosphorylated, together with IRF9, toform the tripartite transcription factor IFN-stimulated factor gene3 (ISGF3), which binds to distinct IFN-stimulated elements ofgenomic DNA for gene transcription (19, 20).Indeed, STAT3 ChIP-seq data displayed peaks in the pro-

motor region, near the transcription start sites of STAT2, IRF7,and IRF9, and a peak in the enhancer region of STAT1. Thesepeaks were significantly reduced after AZD1480 treatment,suggesting that they are direct targets of STAT3 (Fig. 3A). In-creased expression of STAT1, STAT2, and IRF9 was observedwhen STAT3 was knocked down by two different shRNAs inboth TMD8 and OCI-Ly10 cell lines (Fig. 3B). Immunoblot

analysis confirmed that protein levels of STAT1, STAT2, IRF7,and IRF9 were all increased by STAT3 shRNA in these twoABC DLBCL cell lines but not in the control GCB cell lineSUDHL7 (Fig. 3C). In addition, phosphorylation of STAT1 andSTAT2 was remarkably increased in the STAT3 shRNA ex-pressing ABC DLBCL cells (Fig. 3C). To further validate theseresults, we used the constitutively activated form of STAT3(STAT3-C) with activating mutations (A661C and N663C) in theSH2 domain (21). Retroviral expression of STAT3-C in OCI-Ly10 and HBL1 ABC DLBCL cells reduced IRF7 and IRF9expression and completely removed STAT1 phosphorylation(Fig. 3D). Taken together, these data suggest that STAT3 ac-tivity blocks the type I IFN signaling pathway by inhibiting ex-pression of multiple essential signaling components, includingSTAT1, STAT2, IRF7, and IRF9.

Synergism Between STAT3 Inhibition and Lenalidomide in GrowthInhibition of ABC DLBCL. Lenalidomide, an active agent in ABCDLBCL, induces type I IFN response by down-regulation of IRF4and SPIB, which otherwise inhibit IRF7 expression (12, 22). Givenpartial inhibition of IRF4 expression by lenalidomide (12) and thatIRF4 is not a direct target of STAT3, we hypothesized that in-hibition of STAT3 activity augments IFNβ production and syner-gizes with lenalidomide in killing ABC DLBCL cells. To test thishypothesis, we performed an in vitro survival assay in TMD8 andOCI-Ly10 cells when STAT3 shRNA was induced for expression inthe presence of lenalidomide. We used the GCB DLBCL cell lineSUDHL7 as a control. As shown in Fig. 4A, after 3 d of culture,both STAT3 shRNA and lenalidomide significantly reduced cellviability in the two ABC DLBCL cell lines but not in the control.Of note, expression of STAT3 shRNA increased lenalidomide-mediated cytotoxicity in these ABC DLBCL cultures (Fig. 4A).A reduction in viable cells was mainly due to inhibition of cellproliferation (Fig. S3B) although a slight increase in apoptosis wasobserved (Fig. S4). Quantitative PCR analysis confirmed that IFNβexpression was significantly increased in cells that expressedSTAT3 shRNA and were treated with lenalidomide, consistentwith the above survival assay demonstrating a cytotoxic synergismbetween STAT3 shRNA and lenalidomide (Fig. 4B).To further examine the above synergistic effect, we used

ruxolitinib, a clinically used JAK1 and JAK2 inhibitor (23), to

JAK1

MYD88 (L265P)

IRF9

IL-6/IL-10

P

Cell survivalproliferation

NF- BIRF7

P

IFN

STAT3P

STAT3P

TYK2

P

PSTAT1

STAT2

P

PSTAT1

STAT2

IRF9

ISRE

Genes STAT3P

STAT3P

Genes

STAT3P

STAT3P

IRF7IRF9

STAT1STAT2

Cell death

IL6/JAK1/STAT3, PI3K, NF- B signaling cell cycle progression...

Fig. 5. Schematic illustration of gene regulation by STAT3 in ABC DLBCL. JAK1 and STAT3 are activated by autocrine IL-6/IL-10 due to the MYD88 L265Pmutation. STAT3 regulates expression of genes that involve many oncogenic pathways, including NF-κB, PI3K, cell cycle, and itself, signaling to promotecancer cell survival and proliferation while suppressing proapoptotic type I IFN signaling.

Lu et al. PNAS | Published online January 2, 2018 | E503

MED

ICALSC

IENCE

SPN

ASPL

US

Dow

nloa

ded

by g

uest

on

Mar

ch 2

8, 2

021

inhibit STAT3 activity. Immunoblot analysis confirmed dose-dependent inhibition of STAT3 phosphorylation in bothTMD8 and OCI-Ly10 cell lines (Fig. 4C). As expected, weobserved a synergism between ruxolitinib and lenalidomide inkilling these cells (Fig. 4D). Cell-cycle analysis revealed that acombination of the two drugs increased G1 phase population(Fig. 4E), suggesting inhibition of cell proliferation. Moreimportantly, our xenograft analysis in the OCI-Ly10 cell linedemonstrated that cotreatment of ruxolitinib and lenalido-mide caused nearly complete tumor growth inhibition duringthe period of treatment, but the single drug treatmentachieved only partial inhibition (Fig. 4F). Thus, these datasuggest that the ruxolitinib and lenalidomide combination is apotential therapeutic strategy for ABC DLBCL cases.

DiscussionDeregulation of the JAK-STAT signaling pathway, such as con-stitutive activation of STAT3, plays a pathogenic role in manyhematologic malignancies (24–26). In DLBCL, STAT3 is acti-vated in the ABC subtype by IL-6/IL-10 and JAK1 to promotecancer cell survival (3–7). STAT3 activity is also associated with apoor prognosis in DLBCL (27, 28). Here, we conducted genome-wide assessment and established a working model of STAT3 in thepathogenesis of ABC DLBCL (Fig. 5). STAT3 acts as both atranscriptional activator and a suppressor. Gene set enrichmentanalysis revealed that genes regulated by STAT3 are involved inseveral oncogenic signaling pathways, including NF-κB, PI3K/AKT/mTORC1, cell-cycle checkpoint, and STAT3 itself. Notably,STAT3 suppresses expression of STAT1, STAT2, IRF7, and IRF9,all of which are critical transcription factors in the type I IFNpathway. Thus, gene regulation by STAT3 in ABC DLBCL ac-centuates the survival signaling pathways while dampening thelethal type I IFN pathway.Crosstalk between the NF-κB and IL-6/IL-10/JAK1/STAT3

signaling pathways is an oncogenic process in ABC DLBCL (8, 29).Through histone H3 phosphorylation but independent of STAT3,JAK1 up-regulates expression of IRF4 and MYD88, which is re-quired for cancer cell survival (3). The present study revealed thatmany other genes in the NF-κB pathway are STAT3 targets, in-cluding NF-κB2 and TRAF1. Interestingly, NF-κB2 signaling isassociated with MYD88 mutations and promotes developmentof the ABC subtype (30). In a transgenic mouse model, TRAF1 isinvolved in lymphomagenesis mediated by constitutively activatedNF-κB2 (31). These findings suggest a role for the noncanonicalNF-κB pathway in the pathogenesis of ABC DLBCL. Disruptionof oncogenic loops between the NF-κB and JAK1/STAT3 signalingpathways by their small-molecule inhibitors produces the syner-gistic cytotoxicity in ABC DLBCL (3, 4, 32).Several biochemical studies have found a phenomenon of

STAT3-mediated suppression of IFN antiviral responses in im-mune cells (33–36). In ABC DLBCL, the type I IFN signalingpathway, which can be activated by the MYD88 L265P mutation,is proapoptotic to the cancer cells (11) (Fig. 5). Our integratedgenomic analysis elucidates the molecular mechanisms of STAT3in suppression of this lethal pathway in ABC DLBCL: active STAT3prevents the cancer cells from producing IFNβ through inhibition ofIRF7 expression and also suppresses transcription of STAT1,STAT2, and IRF9 (ISGF3 complex) to block IFNβ signaling.

The multilayer suppression of IFNβ signaling by STAT3 is oneof the major mechanisms by which autocrine IL-6/IL-10 signal-ing prevents cancer cell death. In addition, this cytokine sig-naling inhibits IFNβ production through the JAK1-mediatedepigenetic mechanism; that is, JAK1 targets histone H3 to in-duce expression of IRF4 and SPIB, which form a transcriptioncomplex to inhibit IRF7 expression (3, 12).The type I IFN signaling pathway has emerged as an effective

therapeutic target in DLBCL. Recent clinical trials have revealedthat lenalidomide treatment alone or combined with immu-nochemotherapy achieved promising efficacy in DLBCL (37–39).However, remission after a single lenalidomide treatment lastedfor only 6 mo (37), suggesting that combinations of targeted agentsthat inhibit distinct survival pathways will be necessary. Ourfindings demonstrated that STAT3 regulates expression of genesthat involve multiple oncogenic pathways and suppresses genes inthe type I IFN signaling pathway. Inhibition of STAT3 activity byruxolitinib synergizes with lenalidomide in growth inhibition ofABC DLBCL cells in vitro and in a xenograft mouse model.Therefore, this study provides a mechanistic rationale for clinicaltrials of ruxolitinib or a specific JAK1 inhibitor and lenalidomidein ABC DLBCL.

Materials and MethodsFull details of the methods used and data analysis are presented in SI Ma-terials and Methods.

Cell Lines and Culture. All doxycycline-inducible human DLBCL cell lines thatexpress the bacterial tetracycline repressor were engineered as describedpreviously (40). Doxycycline (20 ng/mL) was used to induce the expression ofgenes or shRNAs of interest. All cultures were routinely tested for mycoplasmacontamination.

ChIP-Seq Analysis. ChIP-enriched DNA samples were used to create adapter-ligated libraries for massively parallel sequencing with the Ovation UltralowLibrary System V2 (NuGen Technologies) following the manufacturer’s pro-tocol. ChIP-Seq data are available at https://www.ncbi.nlm.nih.gov/geo/(accession no. GSE106844).

RNA-Seq Analysis. Total RNA was extracted using RNeasy plus mini kit (Qia-gen) according to the manufacturer’s protocol. RNA-seq libraries were pre-pared by using the Illumina TruSeq stranded mRNA LT sample preparationkit (Illumina). Sequencing was performed on Illumina Hiseq 2500 at 50-bplength. RNA-seq data are available at https://www.ncbi.nlm.nih.gov/geo/(accession no. GSE106844).

Xenografts. The xenograft tumormodel of human ABC DLBCL lymphomawasestablished by s.c. injection of OCI-Ly10 cells into nonobese diabetic/severecombined immunodeficiency (NOD/SCID) mice (Jackson Labs). The tumorgrowth was monitored by measuring tumor size in two orthogonal dimen-sions. All animal experiments were approved by the National Cancer In-stitute Animal Care and Use Committee (NCI ACUC) and were performed inaccordance with NCI ACUC guidelines.

ACKNOWLEDGMENTS. This research was supported by the University of Wis-consin, Madison, Start-up Funds and KL2 Scholar Award (UL1TR0000427 andKL2TR000428); the National Cancer Institute (Grant 1R01 CA187299) (to L.R.);the University of Wisconsin, Madison, T32 Hematology Training Award (T32HL07899) (to A.C.D.); the University of Wisconsin Forward Lymphoma Fund(L.L. and S.K.); and the Intramural Research Program of the National CancerInstitute (T.A.W.).

1. Alizadeh AA, et al. (2000) Distinct types of diffuse large B-cell lymphoma identified by

gene expression profiling. Nature 403:503–511.2. Lenz G, Staudt LM (2010) Aggressive lymphomas. N Engl J Med 362:1417–1429.3. Rui L, et al. (2016) Epigenetic gene regulation by Janus kinase 1 in diffuse large B-cell

lymphoma. Proc Natl Acad Sci USA 113:E7260–E7267.4. Lam LT, et al. (2008) Cooperative signaling through the signal transducer and acti-

vator of transcription 3 and nuclear factor-kappaB pathways in subtypes of diffuse

large B-cell lymphoma. Blood 111:3701–3713.5. Ding BB, et al. (2008) Constitutively activated STAT3 promotes cell proliferation and survival

in the activated B-cell subtype of diffuse large B-cell lymphomas. Blood 111:1515–1523.

6. Scuto A, et al. (2011) STAT3 inhibition is a therapeutic strategy for ABC-like diffuse

large B-cell lymphoma. Cancer Res 71:3182–3188.7. Zheng M, et al. (2016) A mix of S and ΔS variants of STAT3 enable survival of activated

B-cell-like diffuse large B-cell lymphoma cells in culture. Oncogenesis 4:e184.8. Staudt LM (2010) Oncogenic activation of NF-kappaB. Cold Spring Harb Perspect Biol

2:a000109.9. Davis RE, et al. (2010) Chronic active B-cell-receptor signalling in diffuse large B-cell

lymphoma. Nature 463:88–92.10. Compagno M, et al. (2009) Mutations of multiple genes cause deregulation of NF-

kappaB in diffuse large B-cell lymphoma. Nature 459:717–721.

E504 | www.pnas.org/cgi/doi/10.1073/pnas.1715118115 Lu et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 2

8, 2

021

11. Ngo VN, et al. (2011) Oncogenically active MYD88 mutations in human lymphoma.Nature 470:115–119.

12. Yang Y, et al. (2012) Exploiting synthetic lethality for the therapy of ABC diffuse largeB cell lymphoma. Cancer Cell 21:723–737.

13. Zhu F, Hwang B, Miyamoto S, Rui L (2017) Nuclear import of JAK1 is mediated by aclassical NLS and is required for survival of diffuse large B-cell lymphoma. Mol CancerRes 15:348–357.

14. Hardee J, et al. (2013) STAT3 targets suggest mechanisms of aggressive tumorigenesisin diffuse large B-cell lymphoma. G3 (Bethesda) 3:2173–2185.

15. Bailey TL, et al. (2009) MEME SUITE: Tools for motif discovery and searching. NucleicAcids Res 37:W202–W208.

16. Zhang Y, et al. (2008) Model-based analysis of ChIP-seq (MACS). Genome Biol 9:R137.17. Mi H, Muruganujan A, Casagrande JT, Thomas PD (2013) Large-scale gene function

analysis with the PANTHER classification system. Nat Protoc 8:1551–1566.18. Krebs DL, Hilton DJ (2000) SOCS: Physiological suppressors of cytokine signaling. J Cell

Sci 113:2813–2819.19. Borden EC, et al. (2007) Interferons at age 50: Past, current and future impact on

biomedicine. Nat Rev Drug Discov 6:975–990.20. Cheon H, et al. (2013) IFNβ-dependent increases in STAT1, STAT2, and IRF9 mediate

resistance to viruses and DNA damage. EMBO J 32:2751–2763.21. Bromberg JF, et al. (1999) Stat3 as an oncogene. Cell 98:295–303.22. Gribben JG, Fowler N, Morschhauser F (2015) Mechanisms of action of lenalidomide in

B-cell non-Hodgkin lymphoma. J Clin Oncol 33:2803–2811.23. Mesa RA, Yasothan U, Kirkpatrick P (2012) Ruxolitinib. Nat Rev Drug Discov 11:

103–104.24. O’Shea JJ, Holland SM, Staudt LM (2013) JAKs and STATs in immunity, immunodefi-

ciency, and cancer. N Engl J Med 368:161–170.25. Chen E, Staudt LM, Green AR (2012) Janus kinase deregulation in leukemia and

lymphoma. Immunity 36:529–541.26. Waldmann TA, Chen J (2017) Disorders of the JAK/STAT pathway in T cell lymphoma

pathogenesis: Implications for immunotherapy. Annu Rev Immunol 35:533–550.

27. Huang X, et al. (2013) Activation of the STAT3 signaling pathway is associated withpoor survival in diffuse large B-cell lymphoma treated with R-CHOP. J Clin Oncol 31:4520–4528.

28. Ok CY, et al. (2014) Clinical implications of phosphorylated STAT3 expression in denovo diffuse large B-cell lymphoma. Clin Cancer Res 20:5113–5123.

29. Rui L, Schmitz R, Ceribelli M, Staudt LM (2011) Malignant pirates of the immunesystem. Nat Immunol 12:933–940.

30. Guo X, et al. (2017) Molecular impact of selective NFKB1 and NFKB2 signaling onDLBCL phenotype. Oncogene 36:4224–4232.

31. Zhang B, Wang Z, Li T, Tsitsikov EN, Ding HF (2007) NF-kappaB2 mutation targetsTRAF1 to induce lymphomagenesis. Blood 110:743–751.

32. Kelly PN, et al. (2015) Selective interleukin-1 receptor-associated kinase 4 inhibitorsfor the treatment of autoimmune disorders and lymphoid malignancy. J Exp Med212:2189–2201.

33. Qing Y, Stark GR (2004) Alternative activation of STAT1 and STAT3 in response tointerferon-gamma. J Biol Chem 279:41679–41685.

34. Costa-Pereira AP, et al. (2002) Mutational switch of an IL-6 response to an interferon-gamma-like response. Proc Natl Acad Sci USA 99:8043–8047.

35. Mahony R, et al. (2017) A novel anti-viral role for STAT3 in IFN-α signalling responses.Cell Mol Life Sci 74:1755–1764.

36. Ho HH, Ivashkiv LB (2006) Role of STAT3 in type I interferon responses. Negative regu-lation of STAT1-dependent inflammatory gene activation. J Biol Chem 281:14111–14118.

37. Wiernik PH, et al. (2008) Lenalidomide monotherapy in relapsed or refractory ag-gressive non-Hodgkin’s lymphoma. J Clin Oncol 26:4952–4957.

38. Vitolo U, et al.; Fondazione Italiana Linfomi (2014) Lenalidomide plus R-CHOP21 inelderly patients with untreated diffuse large B-cell lymphoma: Results of theREAL07 open-label, multicentre, phase 2 trial. Lancet Oncol 15:730–737.

39. Czuczman MS, et al. (2017) A phase 2/3 multicenter, randomized, open-label study oflenalidomide vs investigator’s choice in patients with relapsed or refractory diffuselarge B-cell lymphoma. Clin Cancer Res 23:4127–4137.

40. Ngo VN, et al. (2006) A loss-of-function RNA interference screen for molecular targetsin cancer. Nature 441:106–110.

Lu et al. PNAS | Published online January 2, 2018 | E505

MED

ICALSC

IENCE

SPN

ASPL

US

Dow

nloa

ded

by g

uest

on

Mar

ch 2

8, 2

021