ORIGINAL ARTICLE

Malignant Tregs express low molecular splice forms of FOXP3 in Sezary syndrome

T Krejsgaard1,2, LM Gjerdrum3, E Ralfkiaer3, B Lauenborg1,2, KW Eriksen1,2, A-M Mathiesen1,2, LF Bovin4, R Gniadecki5,C Geisler2, LP Ryder6, Q Zhang7, MA Wasik7, N Ødum1,2 and A Woetmann1,2

1Department of Biology, University of Copenhagen, Copenhagen, Denmark; 2Institute of International Health, Immunology andMicrobiology, University of Copenhagen, Copenhagen, Denmark; 3Department of Pathology, University Hospital of Copenhagen(Rigshospitalet), Copenhagen, Denmark; 4Institute for Inflammation Research, University Hospital of Copenhagen, Copenhagen,Denmark; 5Department of Dermatology, University of Copenhagen, Bispebjerg Hospital, Copenhagen, Denmark; 6Tissue TypingLaboratory, Department of Immunology, University of Copenhagen, Copenhagen, Denmark and 7Department of Pathology andLaboratory Medicine, University of Pennsylvania, Philadelphia, PA, USA

Sezary syndrome (SS) is an aggressive variant of cutaneousT-cell lymphoma. During disease progression, immuno-deficiency develops; however, the underlying molecular andcellular mechanisms are not fully understood. Here, we studythe regulatory T cell (Treg) function and the expression ofFOXP3 in SS. We demonstrate that malignant T cells in 8 of 15patients stain positive with an anti-FOXP3 antibody. Westernblotting analysis shows expression of two low molecular spliceforms of FOXP3, but not of wild-type (wt) FOXP3. The malignantT cells produce interleukin-10 and TGF-b and suppress thegrowth of non-malignant T cells. The Treg phenotype and theproduction of suppressive cytokines are driven by aberrantactivation of Jak3 independent of the FOXP3 splice forms. Incontrast to wt FOXP3, the low molecular splice forms of FOXP3have no inhibitory effect on nuclear factor-jB (NF-jB) activity inreporter assays which is in keeping with a constitutive NF-jBactivity in the malignant T cells. In conclusion, we show that themalignant T cells express low molecular splice forms of FOXP3and function as Tregs. Furthermore, we provide evidence thatFOXP3 splice forms are functionally different from wt FOXP3and not involved in the execution of the suppressive function.Thus, this is the first description of FOXP3 splice forms inhuman disease.Leukemia (2008) 22, 2230–2239; doi:10.1038/leu.2008.224;published online 4 September 2008Keywords: cutaneous T-cell lymphoma; FOXP3; Tregs; Stat; Jak3;NF-kB

Introduction

Mycosis fungoides (MF) and the leukemic variant Sezarysyndrome (SS) are the two major clinical forms of cutaneousT-cell lymphoma (CTCL). The etiology remains poorly under-stood, but occupational exposures, infectious agents and geneticmutations have been proposed as etiological factors. In theinitial phase, which can last a number of years, MF presents asflat erythematous skin patches, resembling chronic inflamma-tion seen in chronic skin infections, psoriasis and eczema. Inthe later stages, MF lesions gradually form plaques and overttumors, and eventually involve lymph nodes and internalorgans.1–5 SS is a particular aggressive variant of CTCL with amean survival of 3 years from the time of diagnosis. The disease

is characterized by the presence of circulating lymphocytes withatypical cerebriform nuclei (Sezary cells) in the skin, lymphnodes and peripheral blood.1

The early skin lesions contain a mixture of cells, includingmalignant T cells that phenotypically resemble normal activatedCD4þ T lymphocytes as well as non-malignant tumor-infiltrat-ing T cells, dendritic cells, macrophages and other inflammatorycells.1–5 During the course of the disease, a shift in thecomposition of infiltrating T cells takes place and the inter-actions between malignant and non-malignant cells change.6

In the early stages of the disease, the infiltrate consists primarilyof non-malignant T helper 1 (TH1) cells and cytotoxic CD8þ

T cells, which appear to control the malignant T cells throughinhibitory cytokines, such as interferon (IFN)-a and IFN-g, andcytotoxicity directed against malignant T cells expressing tumor-associated antigens.7–9 In later stages of the disease, the malig-nant T cells show aberrant hyperactivation of Janus kinase-3(Jak3)10 and signal tranducers and activators of transcription(STAT)11,12 proteins, which in turn trigger survival signals andthe expression of suppressor of cytokine signalling-3 (SOCS3).13

SOCS3 protects the malignant T cells from inhibition byIFNs13 and is a marker of a poor prognosis.14 During diseaseprogression, the normal T-cell compartment is depleted and theimmune system becomes increasingly compromised, leadingto a state of immunodeficiency. Indeed, patients with advancedlesions frequently die of opportunistic infections rather thancomplications from the tumor burden.15 Using an in vitromodel, Berger et al.16 recently provided evidence that CTCLinvolves a malignant proliferation of regulatory T cells(Tregs) with the capacity to inhibit normal T-cell function andpoiesis. To test this hypothesis, several studies have addressedwhether malignant T cells from SS or MF have a regulatoryphenotype in vivo.17–19 The most specific and widely usedmarker for Tregs is FOXP3, a member of the forkhead familyof DNA-binding proteins.20 However, in a previous study,Klemke et al.17 found a decrease in CD4þFOXP3þ cellswithin the dermal lymphomononuclear infiltrate of SS patientswhen compared with control samples, indicating that themalignant T cells do not express FOXP3. Similarly, malignantT cells in MF very rarely express FOXP3 as judged fromimmunohistochemical analysis of tissue sections from MFpatients.17,19

Recently, two low molecular splice forms of FOXP3, desig-nated FOXP3D2 and FOXP3D2D7, have been identified.21,22 Asshown in Figure 1, both splice forms lack exon 2, which is acentral part of the N-terminal proline-rich sequence (repressorregion) that has been shown to be important for FOXP3-medi-ated transcriptional repression.23–25 Moreover, the FOXP3D2D7

Received 21 December 2007; revised 19 June 2008; accepted 11 July2008; published online 4 September 2008

Correspondence: Dr A Woetmann, Department of Biology, Institute ofInternational Health, Immunology and Microbiology, Panum Institute22.5.28, University of Copenhagen, Blegdamsvej 3c, Copenhagen NDK2200, Denmark.E-mail: awoetmann@sund.ku.dk

Leukemia (2008) 22, 2230–2239& 2008 Macmillan Publishers Limited All rights reserved 0887-6924/08 $32.00

www.nature.com/leu

splice form lacks exon 7, which constitute a major part of aleucine zipper domain that is essential for FOXP3 homodimeri-zation.23 However, little is known about the biological functionsof the low molecular splice forms of FOXP3 and no studieshave addressed their role in human disease. Here, we haveinvestigated the expression of FOXP3 in SS, and provide the firstevidence that the malignant T cells selectively express the twolow molecular splice forms of FOXP3 and that the splice formsdiffer functionally from wt FOXP3 in terms of nuclear factor-kB(NF-kB) regulation. Furthermore, we have investigated thesuppressive capacity of the malignant T cells and show thatthey are functional Tregs in vitro. As malignant expression ofFOXP3 splice forms and Treg function is driven by Jak3, ourstudy sheds new light on the function of the Jak3/STAT pathwayin malignant transformation and immunodeficiency in cancer.

Materials and methods

Antibodies and reagentsAntibodies (Abs) against Erk1/2 (K-23) and Jak3 (C-21) were fromSanta Cruz Biotechnology (Santa Cruz, CA, USA). Abs againstStat3 (no. 9132), phospho-Stat5 (no. 9351), NF-kB (no. 3034)and IkB-a (no. 9242) were obtained from Cell SignalingTechnology (Beverly, MA, USA); and the IL-10RA monoclonalAb (mAb) (MAB274) from R&D Systems (Minneapolis, MN,USA). The phospho-Stat3 (Y705) mAb was from NanoTools(Denzlingen, Germany), whereas the Stat5 mAb (610192)was from Becton Dickinson (BD) (Franklin Lakes, NJ, USA).The FOXP3 mAbs 236A/E7 (no. 14-4777) and 150D/E4 (no.14-4774) were from eBioscience (San Diego, CA, USA). TheFITC-conjugated CD25 (F0801) mAb was purchased from Dako(Glostrup, Denmark), whereas the APC-conjugated CTLA-4mAb (555855) together with respective fluorochrome-conju-gated isoform control Abs was obtained from BD. FOXP3, Stat3,Stat5a, Stat5b, Jak3 and non-targeting ON-TARGETplus SMART-pool siRNA were purchased from Dharmacon (Chicago, IL,USA). The Jak3 inhibitor (Jak3I) (WHI-P154) (no. 420104), theLck inhibitor (LckI) (no. 428205), the NF-kB activation inhibitor(NFkBI) (no. 481406) and PD98059 (no. 513000) werepurchased from Calbiochem (San Diego, CA, USA). Finally,tyrphostin Ag1478 and PP1 were from Alexis (Laufelfingen,Switzerland).

PatientsAfter review of histological samples and clinical records, a totalof 15 patients were found diagnosed with SS. The patients were12 males and 3 females with a mean age of 69 years (range44–77). Survival data could be retrieved for five of the cases andthe median survival was 40 months. Biopsies from 11 patients

(diagnosed with SS during the period 1979–2004) weredrawn from the archives at the Department of Pathology atRigshospitalet and at the University of Copenhagen, BispebjergHospital. Primary tumor cells were acquired from peripheralblood of six patients. From two patients, both skin biopsies andprimary tumor cells from peripheral blood could be obtained.Peripheral blood CD3 counts were between 3729 and 33800(mean value of 13 600), CD4/CD8 ratios between 18:1and 386:1 (mean value of 270:1) and percentages of CD4þ ,CD26� cells between 78 and 96% (mean value of 88%).

Cell lines and cell cultureThe malignant T-cell lines SeAx and Sez-4 were establishedfrom peripheral blood of patients diagnosed with SS asdescribed earlier.26,27 The malignant T-cell line, MyLa2000(MF2000), and the non-malignant T-cell lines, MySi andMF1850, were obtained from patients with MF.28–30 The JurkatT-cell line, J-Tag, and the B-cell line, Ramos 2G6.AC10, havepreviously been described.31,32 MF2000, Jurkat and Ramos weregrown in conditional media (RPMI 1640, 2mM L-glutamine,100 mg/ml penicillin/streptomycin all from Sigma, St Louis,MO, USA) supplemented with 10% fetal bovine serum (FBS) (LifeTechnologies, Roskilde, Denmark). SeAx and Sez-4 were grownin conditional media supplemented with 10% pooled humanserum (Blood Bank, State University Hospital, Copenhagen,Denmark) and 103U/ml IL-2 (Proleukin) (Chiron, Emeryville, CA,USA). Finally, MySi and MF1850 were grown in the same mediaas SeAx except that 25ng/ml IL-4 (Leinco, St Louis, MO, USA)was added.

ImmunohistochemistryFollowing microwave heat-induced epitope retrieval, sectionswere incubated for 60min at room temperature with a 1:50dilution of FOXP3 Ab (236A/E7) and then stained using theDAKO Real EnVision Detection System, Peroxidase/DAB,Rabbit/Mouse (K5007) by the Techmate 500 Immunostainer.Scoring of FOXP3þ malignant T cells was performed by two ofthe authors (LMG and ER). FOXP3 labeling of the malignantT cells was scored as negative (no visible staining or positivestaining in o10%), moderately positive (positive staining in10–50% of the malignant T cells) and positive in a majority(450%) of the malignant T cells.

Protein extraction and western blottingProtein extraction and western blotting (WB) were performed asdescribed earlier.10 To ensure equal loading, the total proteinconcentration of each sample was determined by Bio-RadProtein Assay (no. 500-0001; Bio-Rad, Hercules, CA, USA).

Figure 1 Schematic structure of human FOXP3 proteins. Exon organization of predicted structural and functional domains in human wild-type(WT) FOXP3, FOXP3D2 and FOXP3D2D7 proteins. ZnF, zinc finger domain; Zip, leucine zipper domain.

FOXP3 splice forms in malignant T cellsT Krejsgaard et al

2231

Leukemia

Preparation of cytoplasmic and nuclear protein extractions wasmade using the NucBuster protein extraction kit from Novagen,following the manufacturer’s protocol (Novagen, Merck KGaA,Darmstadt, Germany).

Plasmid constructsThe NF-kB-Luc reporter construct were a generous gift from DrMohamed Oukka, and has been described earlier.24 FOXP3splice variants �D2 and �D2D7 were cloned from SeAx cDNAusing PCR with primers containing restriction sites for HindIII inthe forward primer: 50-CCCAAGCTTTGGACAAGGACCCGATG, and XhoI in the reverse primer: 50-CCGCTCGAGGGGGCCAGGTGTAGGGT. The products were ligated into pcDNA3.1expression vector (Invitrogen, Carlsbad, CA, USA). FOXP3 wtwas cloned from an FOXP3-positive non-malignant cell line asdescribed above. All tree inserts were re-cloned into theHA-tagged expression vector pET/HA (a generous gift from DrThomas Mustelin33) using PCR with primers containing restric-tion sites for EcoRI in the forward primer: 50-GGAATTCTTGGACAAGGACCCGATG, and SpeI in the reverse primer: 50-GACTAGTCTCAGGGGCCAGGTGTAGGGT. All PCR reactionswere conducted using the Phusion high-fidelity DNA poly-merase in accordance with the manufacturer’s instruction(Finnzymes, Espoo, Finland). All expression constructs weresequenced bidirectional, and no errors were found.

Transient transfections of siRNATransient transfections were essentially performed as describedearlier34 using 0.5 nmol of the respective ON-TARGETplusSMARTpool siRNAs and 2� 106 cells.

Enzyme-linked immunosorbent assayCells (1� 106 cells per sample) were washed extensively andresuspended in 1ml fresh cytokine-free media in a 12-wellplate. Subsequently, the cells were incubated for 8 or 24 h, withthe given concentrations of inhibitors. Finally, the supernatantswere harvested, and the concentrations of IL-10 and TGF-b1was measured by ELISA using IL-10 (DY217B) or TGF-b1(DY240) Human DuoSet ELISA Development Kits purchasedfrom R&D Systems. For TGF-b1 ELISA, the cells were grown in1% human serum to minimize the background level of TGF-b1.Moreover, supernatants subjected to TGF-b1 ELISA wereactivated prior ELISA analysis in accordance with the manu-facturer’s instructions.

Mixed lymphocyte culture and proliferation assaysPeripheral blood mononuclear cells (PBMCs) were isolated frombuffy coats from healthy donors provided by the UniversityHospital of Copenhagen using Lymphoprep (Axis-ShieldPoC AS, Oslo, Norway) density gradient centrifugation. Then,the PBMCs were washed and resuspended in conditional mediacontaining 10% human serum. Stimulator cells were irradiatedfor 10min at 3000 rad and mixed 1:1 with responder cells inthe bottom chamber of a 12-well plate with transwell inserts(pore size 0.4mM) (Greiner Bio-One, Frickenhausen, Germany)in the presence or absence of SeAx cells in the upper chamberand anti-IL-10RA (500mg/ml) or anti-TGF-b (0.1 mg/ml) as given.The cells were cultured for 5 days with the addition of 7 mCi[3H]thymidine to each well 20 h prior harvest. Finally, thesamples from the bottom chamber were transferred to 96-wellround-bottomed plates, harvested onto glass fiber filters and the[3H]thymidine incorporation was measured in a TopCount

scintillation counter (PerkinElmer, Waltham, MA, USA). Theproliferation was expressed as mean counts per minute (CPM) ofgiven numbers of replicate cultures. Similarly, non-malignantT cells from a CTCL patient (MF1850) were cultured in condi-tional media with 10% human serum and 1000U/ml IL-2 in thebottom chamber with or without SeAx in the upper chamber.The SeAx cells were either transfected or not transfected withnon-targeting or FOXP3 siRNA 24h before co-culture. After3 days of culture, 7 mCi of [3H]thymidine was added to eachwell and incorporation determined as described.35

RNA purification and RT-PCRTotal cellular mRNA was isolated using the RNeasy RNApurification kit from Qiagen with the addition of on-columnDNase digestion to avoid DNA contamination as described bythe manufacturer (Qiagen, Hilden, Germany). Purified RNA wasreverse transcribed using murine Moloney leukemia virusreverse transcriptase (M-MLV, Invitrogen). The resulting cDNAwas amplified by PCR using recombinant Taq DNA polymerase(New England Biolabs, Beverly, MA, USA) with the followingprimers: GADPH forward: 50-CCATGGAGAAGGCTGGGG,GADPH reverse: 50-CAAAGTTGTCATGGATGACC, CTLA-4forward: 50-CTCAGCTGAACCTGGCTACC, CTLA-4 reverse:50-CTTCAGTCACCTGGCTGTCA GITR forward: 50-GAGTGGTGCATGTGTGT, GITR reverse: 50-TCTGTCCAAGGTTTGCAGTG, FOXP3 forward: 50-CATGATCAGCCTCACACCAC,FOXP3 reverse: 50-CCACTTGCAGACACCATTTG, Galectin-10forward: 50-CTACCCGTGCCATACACAGA, Galectin-10 reverse:50-GTTCATGACCACACGACGAC. PCR programme as follows:94 1C for 2min, 28 cycles of 94 1C for 30 s, at 56 1C for 30 s, at72 1C for 30 s, and finally at 72 1C for 10min using the Taq DNApolymerase from New England Biolabs (no. M0267). The sampleswere analyzed in a 2% agarose gel.

Flow cytometryFor surface staining, cells were harvested, washed in ice-coldFACS buffer (phosphate-buffered saline (137mM NaCl, 10mM

phosphate, 2.7mM KCL, pH 7.4), 5% FBS and 0.1% sodiumazide) and stained with 5ml FITC-conjugated anti-CD25or isotype control Abs for 30min at 4 1C in the dark.For intracellular staining, cells were harvested, washed inphosphate-buffered saline and fixed and permeabilized usingBD Cytofix/Cytoperm Fixation/Permeabilization Solution Kit(554714) according to the manufacturer’s protocol. Then, theywere stained with 5ml APC-conjugated anti-CTLA-4 or isotypecontrol Abs for 30min at 4 1C in the dark. After final washing,the cells were resuspended in FACS buffer and analyzed on aFACSCalibur using CellQuest software (BD).

Luciferase assayJurkat J-TAg cells were transiently co-transfected with 1.0 mgNF-kB-Luc reporter construct, 3.0 mg of a pcDNA3.1 expressionvector either empty or encoding FOXP3- wt, �D2 or �D2D7,respectively, and 0.5mg Renilla luciferase control vector(pRL-CMV; Promega, Madison, WI, USA) with the DMRIE-Ctransfection reagent, according to the manufacturer’s protocol(no. 10459-014, Invitrogen). At 24 h after transfection, cellswere left unstimulated or stimulated with 10 ng/ml flagellin(Alexis) for an additional 24 h. The cells were harvested, andthe luciferase activities were determined by using the Dual-Luciferase Reporter Assay system (Promega) according to themanufacturer’s instruction. The co-expressed Renilla luciferaseactivity was used for normalization of transfection efficiency.

FOXP3 splice forms in malignant T cellsT Krejsgaard et al

2232

Leukemia

EthicsThe study was approved by the Local Ethics Committee (journalno. 01 284225) and the Danish Data Protection Agency (journalno. 2005-41–5930).

Results

Malignant T cells express FOXP3 splice formsTo address whether malignant T cells display a Treg phenotypein vivo, we performed immunohistochemical stainings of skinspecimens from SS patients with an anti-FOXP3 Ab (236A/E7). Intotal, malignant T cells in 4 of 11 patients displayed a Tregphenotype as judged from FOXP3 reactivity in the skin. Asexemplified in Figure 2a, the majority of malignant T cellsdisplayed a strong nuclear staining in the FOXP3-positivepatients. In contrast, cytoplasmic staining was only observedin a minority of malignant T cells and the stroma was overallnegative (Figure 2a and data not shown). We also found thatthe leukemic T-cell lines, SeAx and Sez-4, established fromtwo different patients with SS, displayed nuclear expression ofFOXP3 (Figure 2b). To verify the specificity of the Ab, WB wasperformed on cell lysates from a FOXP3-positive cell line, aFOXP3-negative cell line and the SeAx cell line. As expected, aprotein with a molecular weight of approximately 48 kDa equalto the molecular weight of wild-type (wt) FOXP3 was identifiedin the lysate from the FOXP3-positive cell line but not in thelysate from the FOXP3-negative cell line (Figure 3a). Similarly,wt FOXP3 was detected in cell lysates from FOXP3-negativeJurkat cells transfected with wt FOXP3 but not Jurkat cellstransfected with a vector control (Figure 3b). Surprisingly, wtFOXP3 was not detectable in the SeAx cells (Figures 3a and b).Thus, probing with an Ab (150D/E4) that selectively reacts withwt FOXP3 (exon 2) showed that the SeAx cells did not expressdetectable levels of wt FOXP3 (Figure 3b, lower panel). How-ever, two proteins of lower molecular weights were identifiedin the lysate from SeAx (but not in FOXP3-negative) cells bythe FOXP3 Ab 236A/E7, which recognizes all known forms ofFOXP3 (Figures 3a and b). To address whether the two low-molecular-weight proteins were identical to the two recentlyidentified FOXP3 splice forms, FOXP3D2 and FOXP3D2D7

(Figure 1), we performed WB on cell lysates from SeAx cells andFOXP3-negative Jurkat cells transfected with a vector control,wt FOXP3 or one or both of the two low molecular splice formsof FOXP3. As shown in Figure 3c, the FOXP3 Ab (236A/E7)recognized wt FOXP3 and both the FOXP3 splice forms whenexpressed in Jurkat cells. The Ab also recognized the two low-molecular-weight proteins in cell lysates from the SeAx cell line(Figure 3c). The two low-molecular-weight proteins in the SeAxcells had molecular weights similar to that of the FOXP3D2 andFOXP3D2D7 splice forms expressed in Jurkat cells, but clearlylower than that of wt FOXP3 (Figure 3c). Identical results wereobtained with the Sez-4 cell line (data not shown). Furthersupporting the notion that the two proteins are low molecularsplice forms of FOXP3, transfection of SeAx cells with FOXP3siRNA induced a selective knock down of both bands, whereasthe expression of Stat3 remained unaffected (Figure 3d). Finally,both splice forms could be cloned from the SeAx and Sez-4 celllines (data not shown) and, notably, no mutations were iden-tified by sequence analyses (data not shown). As in the immuno-histochemical analyses (Figure 2b), the low molecular spliceforms of FOXP3 were located within the nuclear fraction of SeAxand SeZ-4 cells (Figure 3e and data not shown). Collectively,these results provide evidence that the SeAx and Sez-4 cell linesselectively express the two low molecular splice forms, FOXP3D2and FOXP3D2D7, of FOXP3. In general, the FOXP3D2D7 spliceform was expressed at higher levels when compared with theFOXP3D2 splice form, but the expression of both splice formsvaried to some degree depending on cell culture conditions(Figures 2b, 3 and 4 and data not shown). The expression of theFOXP3D2 splice form was very labile and seemed to beparticularly sensitive to the culture conditions. To address whetherprimary malignant T cells in blood of SS patients also express thelow molecular splice forms of FOXP3, cell lysates from peripheralblood of six SS patients were subjected to WB. It can be noted thattwo of these patients had also been included in the immunohis-tochemical analysis and found to be FOXP3-negative. As shownin Figure 3f, four of the patients expressed the two low molecularsplice forms of FOXP3 in peripheral blood, whereas no expressionof wt FOXP3 was detected using either the 236A/E7 or 150D/E4Ab. Confirming our earlier findings, no expression of FOXP3 wasobserved in peripheral blood cells of the two patients found to be

Figure 2 FOXP3 expression in a patient with SS and malignant T-cell lines. (a) Skin biopsy from an SS patient with a dense, epidermotropicinfiltrate of atypical cells (HE, � 200) forming classic Pautrier abscesses (HE, � 600). Immunohistology for FOXP3 (236A/E7) shows a strongnuclear staining in a majority of the malignant cells (FOXP3, �200 and � 600). (b) Immunohistochemical stainings for FOXP3 at �200 and�400 magnification of SeAx and Sez-4 cells.

FOXP3 splice forms in malignant T cellsT Krejsgaard et al

2233

Leukemia

negative in the immunohistochemical analysis (data not shown).Expectedly, STAT3 and STAT5 were activated (as judged from thepresence of tyrosine phosphorylated proteins) in peripheral bloodof all patients, although to varying degrees (Figure 3f and datanot shown). In conclusion, 8 of 15 patients exhibited a Tregphenotype as judged from FOXP3 reactivity in the skin or blood.

Expression of FOXP3 splice forms is driven by Jak3 andSTAT5Because aberrant activation of the Jak3/STAT signaling pathwayis a hallmark of CTCL and seems to drive the malignant

transformation,10,34,36,37 we asked whether the expression of theFOXP3 splice forms was driven by Jak3. Accordingly, SeAx cellswere treated with a Jak3 inhibitor (Jak3I), an inhibitor of the p38MAP kinase (PD98059) or two Src kinase inhibitors (PP1, LckI) priorto the analysis of FOXP3 expression. As expected, the Jak3 inhibitorinduced a dose-dependent inhibition of the constitutive STAT3activity, whereas the other inhibitors did not (Figure 4a). A selectiveinhibition of FOXP3 was also observed in cells treated with the Jak3inhibitor as compared with cells treated with vehicle or the otherinhibitors (Figure 4a), indicating that Jak3 has an important functionin the expression of the low molecular FOXP3 splice forms. Tofurther substantiate this conclusion, SeAx cells were transfectedwith FOXP3, Jak3, STAT3 or STAT5a/b siRNAs prior to analysesof FOXP3 expression. Jak3 siRNA almost completely knockeddown Jak3 expression, whereas other siRNAs had no effect on theexpression of Jak3 (Figure 4b). Similarly, FOXP3 siRNA blockedthe expression of the FOXP3 splice forms (Figure 4b). Importantly,the expression of the FOXP3 splice forms was also stronglyinhibited by Jak3 siRNA when compared with the non-targetingsiRNA control (Figure 4b), supporting that Jak3, indeed, has animportant function in the expression of the FOXP3 splice forms.IL-2-driven expression of FOXP3 has previously been shown to bemediated through STAT5.38 In accordance, we found that down-regulation of STAT5 expression severely inhibited the expression ofthe FOXP3 splice forms (Figure 4b). However, STAT3 siRNA,which almost completely blocked STAT3 expression, had noinhibitory effect of FOXP3 expression (Figure 4b). On the contrary,STAT3 inhibition repeatedly triggered an enhanced expression ofFOXP3 (Figure 4b), which is in keeping with recent observationsthat STAT3 can function as a repressor of FOXP3 expression.38,39

Malignant T cells spontaneously produce IL-10 andTGF-b and suppress non-malignant T cellsIt has been suggested that IL-10 and TGF-b have a key functionin the immune evasion by malignant T cells and development of

Figure 3 Nuclear expression of two low molecular splice forms ofFOXP3 in malignant T cells. (a) WB analyses of FOXP3 and Stat3expressions in a malignant T-cell line established from a patientwith SS (SeAx), a wt FOXP3-negative (FOXP3 WT�) cell line and a wtFOXP3-positive (FOXP3 WTþ ) cell line. (b) Total lysates from SeAx orJurkat cells transfected with empty vector (pEF-HA) or wt FOXP3(HA-FOXP3-WT) were analyzed by WB using either the antibody236A/E7, which recognizes all known forms of FOXP3, or 150D/E4,which recognizes only wt FOXP3 (exon 2). (c) Jurkat cells weretransiently transfected with empty vector (pEF-HA) or the respectivevector encoding HA-tagged wt FOXP3 (HA-FOXP3-WT), FOXP3D2(HA-FOXP3D2) and/or FOXP3D2D7 (HA-FOXP3D2D7) and culturedfor 24 h. Then, the transfected Jurkat and SeAx cells were lysed, andthe FOXP3 expression was analyzed by WB using an anti-FOXP3antibody. (d) SeAx cells were transiently transfected with non-targeting(NT) or FOXP3-specific siRNA and grown in cytokine-free media for48 h. Subsequently, the cells were lysed and the expressions of FOXP3and Stat3 were determined by WB. (e) Lysates from cytoplasmic (Cyt)and nuclear (Nuc) fractions of SeAx cells were analyzed by WB usingFOXP3-, NF-kB- and IkB-a-specific antibodies. (f) Jurkat T cells, SeAxcells and cells obtained from peripheral blood of four patients (P1–P4)diagnosed with SS were lysed and the lysates analyzed by WB.

Figure 4 Jak3 and Stat5 induce the expression of FOXP3, whereasStat3 inhibits FOXP3 expression. (a) SeAx cells were cultured in thepresence of low (100mg/ml) or high (250mg/ml) concentrations ofJak3I, PP1 (20mM), LckI (10mM), PD98059 (10mM) or vehicle (�) for6 h. Next, the cells were lysed, and the expressions of FOXP3, pYStat3and Stat3 were determined by WB. (b) SeAx cells were transientlytransfected with FOXP3, non-targeting (NT), Stat3, Jak3 or a pool ofStat5a and Stat5b (Stat5) siRNAs. Subsequently, they were cultured for48 h in cytokine-free media and total cell lysates analyzed by WB.

FOXP3 splice forms in malignant T cellsT Krejsgaard et al

2234

Leukemia

immunodeficiency in CTCL.40,41 As shown in Figure 5a, theSeAx cells spontaneously produced IL-10 and TGF-b, whereasnon-malignant T cells did not. Inhibition of Jak3 activityprofoundly reduced the spontaneous production of IL-10,whereas an inhibitor of the epidermal growth factor receptor(Ag1478) had no effect (Figure 5a, left). Similarly, down-regulation of Jak3 expression by siRNA clearly reduced thespontaneous IL-10 production (Figure 5b). The spontaneousTGF-b production was also reduced by the Jak3 inhibitor,although to a lesser extent (Figure 5a, right). Because FOXP3splice variants might have a role in the development andfunction of Tregs, we asked whether FOXP3 knockdown bysiRNA inhibited the cytokine production of the SeAx cells. Asshown in Figure 5b, FOXP3 siRNA had no detectable effect ofthe spontaneous IL-10 production. In contrast, STAT3 and, tosome extent, STAT5 siRNA partially inhibited the spontaneousIL-10 production (Figure 5b).To address whether malignant T cells from SS patients

function as Tregs, SeAx cells were co-cultured with non-malignant T cells. Figures 6a and b show co-culture experimentsbetween SeAx cells and non-malignant T cells in a semi-permeable transwell system. As shown, the SeAx cells inhibited

IL-2-driven proliferation of non-malignant T cells from a CTCLpatient (Figure 6a) and a mixed lymphocyte culture betweenhealthy donors by approximately 40% (Figure 6b). Knock downof FOXP3 splice forms by siRNA had no effect on thesuppression (Figure 6a), indicating that the FOXP3 splice formsare not involved in the execution of the Treg function of themalignant T cells. Direct co-culture between irradiated (to avoidproliferation) SeAx cells and non-malignant T cells did nottrigger higher levels of suppression (data not shown), suggestingthat the suppression is mediated by soluble factors rather than bydirect cell contact. In support of this possibility, an IL-10RAblocking mAb and an IL-10 neutralizing mAb inhibited the Tregfunction of the SeAx cells (Figure 6b and data not show).Because an anti-TGF-b Ab had little effect on the suppression(Figure 6b), it seems that the suppression is mediated primarilyby IL-10, which is in keeping with our observation thatexogenous IL-10 inhibit an MLC by 40–50% (data not shown).The addition of exogenous IL-2 increased the proliferation ofnon-malignant T cell in both cultures with and without SeAx,but slightly inhibited the level of suppression from 54% incultures without IL-2 to 37% in cultures with excess of IL-2 (datanot shown).

Figure 5 Jak3-dependent expressions of IL-10 and TGF-b. (a) Non-malignant (MySi) and malignant (SeAx) T cells were cultured in cytokin-freemedia containing Jak3I (50 mg/ml), Ag1478 (200ng/ml) or vehicle (�) for 8 or 24 h. Then, the supernantans were harvested, and the IL-10 or TGF-bproduction was analyzed by ELISA. (b) SeAx cells were transfected with non-targeting (NT), FOXP3, Jak3, Stat3 or a pool of Stat5a and Stat5bsiRNAs and cultured for 48 h in cytokine-free media. Subsequently, the supernatants were harvested, and the IL-10 production was determinedby ELISA. The IL-10 production is presented as percentage IL-10 production relative to SeAx cells transfected with NT siRNA. The data are from five(a) and four (b) independent experiments, respectively, performed with duplicate cultures. Error bars represent s.e.m.

FOXP3 splice forms in malignant T cellsT Krejsgaard et al

2235

Leukemia

Malignant T cells display surface expression of CD25but not of CTLA-4In addition to FOXP3, a number of molecules have been linkedto the Treg phenotype. In particular, not only high expression ofthe IL-2R a-chain (CD25) but also CTLA-4, GITR, andintracellular expression of Galectin-10 have been associatedwith Treg phenotype and function.42,43 As shown pre-viously26,37 and confirmed here, SeAx cells express high levelsof CD25 (Figure 7c). CTLA-4, which has a critical function incell-mediated suppression,44 was detected at the mRNA level(Figure 7a) and by intracellular staining (Figure 7d), but nosignificant surface expression could be detected even aftertreatment with phorbol ester and ionomycin (data not shown).Similarly, GITR was detected at the mRNA level (Figure 7a) butbarely detectable at the protein level (data not shown). Incontrast, Galectin-10 mRNA was not detected in either the SeAxor Sez-4 cell line (Figure 7b).

FOXP3 wt and splice forms have different effects onNF-kB activity in luciferase assaysThe NF-kB signal pathway has been shown to be constitutivelyactive and essential for the protection from apoptosis inperipheral blood tumor cells and T-cell lines from patients with

SS.45 In accordance, we found that NF-kB exhibited constitutivenuclear localization in the SeAx cell line (Figure 3e) and that aninhibitor of NF-kB transcriptional activation46 (NF-kBI) induceddose-dependent growth arrest in the SeAx cells (Figure 8a).Given previous findings that wt FOXP3 inhibits NF-kB activity,24

it was interesting to address whether the low molecular spliceforms of FOXP3 also inhibit NF-kB transcriptional activity.Therefore, Jurkat T cells were transfected with a vector control,FOXP3 wt, FOXP3D2 or FOXP3D2D7 prior to the measurementof flagellin-induced NF-kB activity using a luciferase reporterconstruct. As shown in Figure 8b, flagellin triggered a profoundNF-kB-mediated transcriptional activity as judged from theincrease in luciferase activity. FOXP3 wt inhibited the flagellininduced increase in NF-kB activity. In contrast, the FOXP3splice forms had little or no inhibitory effect on the flagellin-induced transcriptional activity of NF-kB (Figure 8b, upperpanel) despite comparable levels of mRNA expression as wtFOXP3 (Figure 8b, lower panel). Similar effects of FOXP3 wt andsplice forms were seen on IL-1b-induced NF-kB reporter activity(data not shown).

Discussion

In this study, we provide the first evidence that malignant T cellsfrom SS selectively express low molecular splice forms ofFOXP3 in vivo. Furthermore, we show that the malignant T cellscan function as Tregs in vitro. Thus, the malignant T cells

Figure 6 SeAx cells suppress the proliferation of non-malignant Tcells. (a) 1� 104 non-malignant T cells from a CTCL patient (MF1850)were cultured in media containing IL-2 for 4 days (1000U/ml) in thebottom chamber of a transwell system with non-malignant T cells orSeAx cells in the upper chamber as given. The SeAx cells were eithertransfected or not transfected with non-targeting (NT) or FOXP3 siRNA24h before co-culture. (b) A total of 5�103 responder and stimulator(irradiated) peripheral blood mononuclear cells (PBMCs) were mixed1:1 in the lower chamber of a transwell system and cultured for 5 dayswith SeAx cells or without (mixed lymphocyte culture, MLC) SeAxcells in the upper chamber in the presence or absence of anti-IL-10RA(500mg/ml) or anti-TGF-b (100mg/ml) antibodies. (a and b) Twentyhours prior harvest, [3H]thymidine was added to each well, and the[3H]thymidine incorporation into the non-malignant T cells in thebottom wells was measured. The proliferation is expressed as meancounts per minute (CPM) and the error bars represent s.d. Theexperiments (a and b) were performed with eight replicate cultures andare representative of three and two independent experiments,respectively.

Figure 7 SeAx cells exhibit a regulatory phenotype. (a) RT-PCRanalysis of mRNA levels of FOXP3, CTLA-4, GITR and GADPH asinternal control in SeAx and control cell lines. (b) RT-PCR analysis ofmRNA levels of galectin-10 (Gal-10) and GADPH in SeAx, Sez-4,MF2000 and peripheral blood mononuclear cells (PBMCs) as positivecontrol. (c) SeAx cells were stained with an anti-CD25-FITC (open) orFITC-conjugated isotype control (black) antibodies (Abs) and analyzedby flow cytometry. Data are representative of three independentexperiments. (d) SeAx cells were permeabilized and then stained withanti-CTLA-4-APC (open) or APC-conjugated isotype control Abs(black) and analyzed by flow cytometry. Data are representative oftwo independent experiments.

FOXP3 splice forms in malignant T cellsT Krejsgaard et al

2236

Leukemia

suppressed cytokine-driven proliferation of non-malignantT cells isolated from CTCL patients. Similarly, the malignantT cells inhibited allo-antigen-driven proliferation of T cellsfrom healthy donors. Our results show that the suppressive effectwas mediated at least partly by soluble factors. In accordance,the malignant T cells spontaneously produced immuno-suppressive cytokines such as IL-10 and TGF-b, and thesuppressive effect was inhibited by IL-10RA blocking andIL-10 neutralizing Abs.

Our findings that some malignant T cells from SS patientsexpress FOXP3 splice forms are in apparent contrast to a recentreport by Klemke et al.17 However, Klemke et al. did notspecifically investigate the expression of the splice forms.Moreover, as we only find FOXP3 expression in a subpopulationof patients, it is likely that variations between the cohortsof patients might explain our different observations. From ourstudy, it was not possible to assign FOXP3 expression to certainsubgroups of patients, and future investigations are needed toelucidate the frequency of FOXP3-positive patients and whetherthe expression of FOXP3 splice forms is associated with clinicalsubgroups of SS.

Although the SeAx cell line was reported to be FOXP3negative,17 we have been able to clone the FOXP3 splice formsfrom this cell line and find it to be FOXP3 positive inimmunohistochemistry (Figure 1b), WB (Figure 3) and RT-PCR(Figure 7). This discrepancy could be explained by differences inSeAx sublines, in assay sensitivity, and to some degree, byvariations in cell culture conditions between the two studies, aswe find that the level of FOXP3 expression in SeAx cells isinfluenced by the culture conditions.

Our findings that malignant T cells selectively expressingFOXP3 splice forms function as Tregs suggest that the expressionof splice forms (like wt) in combination with CD25 is a maker ofa Treg phenotype. However, siRNA-mediated downregulationof the FOXP3 splice forms did not inhibit cytokine production orTreg function. Thus, our data suggest that the FOXP3 spliceforms might be involved in the differentiation of malignantTreg cells but not in the execution of the suppressive effect.The observations that the expression of the FOXP3 spliceforms relied on Jak3/STAT5 but not on STAT3, whereas cytokineproduction was partly dependent on STAT3, support thisconclusion.

One of the main questions, which arise from this study, is whythe malignant T cells selectively express the low molecularsplice forms of FOXP3 and not wt FOXP3. In some experimentalmodels, wt FOXP3 has been shown to have an anti-proliferativeeffect.22,47 However, downregulation of the FOXP3 splice formshad no effect on the spontaneous proliferation of the SeAx cells(Supplementary Figure 1), suggesting that the splice forms haveno repressive effect on malignant proliferation. In support of thisconclusion, Allan et al.21 showed that FOXP3D2 had little anti-proliferative effect when expressed in transduced T cells relativeto wt FOXP3. It has previously been reported that wt FOXP3inhibits the transcriptional activity of NF-kB.24 However, NF-kBis constitutively activated in malignant T cells from SS patientswhere it has an essential function by inhibiting apoptosis.45

Therefore, it was crucial whether the low molecular splice formsof FOXP3 also inhibited NF-kB transcriptional activity. Expect-edly, wt FOXP3 strongly inhibited NF-kB transcriptional activityin Jurkat cells, whereas the FOXP3 splice forms did not. Thisfinding is in agreement with previous rapports identifying exon 2as a central part of the transcriptional repressor domain inFOXP323–25 (Figure 1). Thus, our findings suggest that selectiveexpression of the FOXP3 splice forms may protect the malignantT cells from an anti-proliferative effect of wt FOXP3 whileretaining the anti-apoptotic effect of NF-kB.

This is the first investigation of FOXP3 splice forms in ahuman disease. Given present and previous findings showingthat wt FOXP3 and splice forms appear to have overlapping butpartly different functions, our observation of a selectiveexpression of FOXP3 splice forms in SS might have implicationsfor other studies of FOXP3 expression in tissue samplesand blood.21,22,25 Thus, most immunohistochemical and flowcytometric studies of FOXP3 expression do not discriminate

Figure 8 wt FOXP3 but not FOXP3D2 or FOXP3D2D7 inhibits NF-kBactivity. (a) SeAx cells were cultured with given concentrations of NF-kBI, Ag1478 or vehicle (�) for 72 h in 96-well round-bottomed plates.At 20 h prior harvest, 1mCi [3H]thymidine was added to each culture.Subsequently, [3H]thymidine incorporation was measured, and theproliferation was expressed as mean counts per minute (CPM) of sixreplicate cultures. Data are representative of three independentexperiments. (b) Jurkat cells were transiently co-transfected with anNF-kB-luciferase reporter construct, a Renilla luciferase control vectorand a pcDNA3.1 expression vector either empty or encoding FOXP3wt, FOXP3D2 or FOXP3D2D7. At 24 h after transfection, the cellswere left unstimulated, or stimulated with 10ng/ml flagellin for anadditional 24 h before harvest. Then, the mRNA levels of FOXP3 andGADPH were analyzed by RT-PCR (lower panel) and the luciferaseactivities were determined (upper panel). The co-expressed Renillaluciferase activity was used for the normalization of transfectionefficiency, and fold induction was calculated relative to respective un-stimulated samples. The experiment was performed with triplicatecultures and is representative of five independent experiments. Errorbars represent s.e.m.

FOXP3 splice forms in malignant T cellsT Krejsgaard et al

2237

Leukemia

between wt FOXP3 and the low-molecular-weight splice forms.Much important information on the function and clinicalimportance of FOXP3 versus splice forms awaits a bettercharacterization of most anti-FOXP3 Abs used in human studies.

In conclusion, we provide the first evidence that malignant Tcells from SS patients selectively express FOXP3 splice formsand that the splice forms differ functionally from wt FOXP3 interms of the regulation of NF-kB activity. As malignantexpression of FOXP3 splice forms and Treg function is drivenby Jak3, our study adds to the understanding of the role of theJak3/STAT signaling pathway in malignant transformation,immune evasion and immunodeficiency in cancer.

Acknowledgements

This work was supported by grants from The University ofCopenhagen, The Scientific Faculty, The Danish ResearchCouncils, The Foundation of 17-12-1981, The Novo NordicFoundation, The Danish Cancer Society, Neye Fonden, TheLundbeck Foundation, The Leo Pharma Foundation, and TheNational Cancer Institute-CA89194 (MA Wasik). We thank KeldKaltoft (Arhus University and CellCure Arhus, Denmark) for thegenerous gift of MF, MySi and SeAx cell lines. Approval wasobtained from the University of Copenhagen for these studies.Informed consent was provided according to the Declaration ofHelsinki Principles. The project part concerning the establishmentand study of CTCL cell lines by Dr Keld Kaltoft has been approvedby ‘Den videnskabsetiske Kommite i Arhus Amt’ (The science-ethical committee in Arhus County).

Conflict of interestThe authors declare no competing financial interests.

References

1 Berger CL, Mariwalla K, Girardi M, Edelson RL. Advances inunderstanding the immunobiology and immunotherapy of cuta-neous T-cell lymphoma. Adv Dermatol 2004; 20: 217–235.

2 Dummer R. Future perspectives in the treatment of cutaneousT-cell lymphoma (CTCL). Semin Oncol 2006; 33: S33–S36.

3 Willemze R, Jaffe ES, Burg G, Cerroni L, Berti E, Swerdlow SHet al. WHO-EORTC classification for cutaneous lymphomas.Blood 2005; 105: 3768–3785.

4 Kim EJ, Hess S, Richardson SK, Newton S, Showe LC, Benoit BMet al. Immunopathogenesis and therapy of cutaneous T celllymphoma. J Clin Invest 2005; 115: 798–812.

5 Bagot M, Boumsell L, Bensussan A. Immunopathogenesis ofcutaneous T-cell lymphomas. Hematol Oncol Clin North Am2003; 17: 1313–1317, vii.

6 Bagot M, Nikolova M, Schirm-Chabanette F, Wechsler J, BoumsellL, Bensussan A. Crosstalk between tumor T lymphocytes andreactive T lymphocytes in cutaneous T cell lymphomas. Ann N YAcad Sci 2001; 941: 31–38.

7 Lee BN, Duvic M, Tang CK, Bueso-Ramos C, Estrov Z, Reuben JM.Dysregulated synthesis of intracellular type 1 and type 2 cytokinesby T cells of patients with cutaneous T-cell lymphoma. Clin DiagnLab Immunol 1999; 6: 79–84.

8 Vowels BR, Lessin SR, Cassin M, Jaworsky C, Benoit B, Wolfe JTet al. Th2 cytokine mRNA expression in skin in cutaneous T-celllymphoma. J Invest Dermatol 1994; 103: 669–673.

9 Linnemann T, Tumenjargal S, Gellrich S, Wiesmuller K, Kaltoft K,Sterry W et al. Mimotopes for tumor-specific T lymphocytes inhuman cancer determined with combinatorial peptide libraries.Eur J Immunol 2001; 31: 156–165.

10 Krejsgaard T, Vetter-Kauczok CS, Woetmann A, Lovato P, LabudaT, Eriksen KW et al. Jak3- and JNK-dependent vascular endothelialgrowth factor expression in cutaneous T-cell lymphoma. Leukemia2006; 20: 1759–1766.

11 Zhang Q, Nowak I, Vonderheid EC, Rook AH, Kadin ME,Nowell PC et al. Activation of Jak/STAT proteins involved insignal transduction pathway mediated by receptor for interleukin 2in malignant T lymphocytes derived from cutaneous anaplasticlarge T-cell lymphoma and Sezary syndrome. Proc Natl Acad SciUSA 1996; 93: 9148–9153.

12 Nielsen M, Kaltoft K, Nordahl M, Ropke C, Geisler C, Mustelin Tet al. Constitutive activation of a slowly migrating isoform of Stat3in mycosis fungoides: tyrphostin AG490 inhibits Stat3 activationand growth of mycosis fungoides tumor cell lines. Proc Natl AcadSci USA 1997; 94: 6764–6769.

13 Brender C, Nielsen M, Kaltoft K, Mikkelsen G, Zhang Q, Wasik Met al. STAT3-mediated constitutive expression of SOCS-3 incutaneous T-cell lymphoma. Blood 2001; 97: 1056–1062.

14 Krishnadasan R, Bifulco C, Kim J, Rodov S, Zieske AW, VanasseGJ. Overexpression of SOCS3 is associated with decreased survivalin a cohort of patients with de novo follicular lymphoma. Br JHaematol 2006; 135: 72–75.

15 Axelrod PI, Lorber B, Vonderheid EC. Infections complicating mycosisfungoides and Sezary syndrome. JAMA 1992; 267: 1354–1358.

16 Berger CL, Tigelaar R, Cohen J, Mariwalla K, Trinh J, Wang N et al.Cutaneous T-cell lymphoma: malignant proliferation of T-regula-tory cells. Blood 2005; 105: 1640–1647.

17 Klemke CD, Fritzsching B, Franz B, Kleinmann EV, Oberle N,Poenitz N et al. Paucity of FOXP3+ cells in skin and peripheralblood distinguishes Sezary syndrome from other cutaneous T-celllymphomas. Leukemia 2006; 20: 1123–1129.

18 Tiemessen MM, Mitchell TJ, Hendry L, Whittaker SJ, Taams LS,John S. Lack of suppressive CD4+CD25+FOXP3+ T cells inadvanced stages of primary cutaneous T-cell lymphoma. J InvestDermatol 2006; 126: 2217–2223.

19 Gjerdrum LM, Woetmann A, Odum N, Burton CM, Rossen K,Skovgaard GL et al. FOXP3+ regulatory T cells in cutaneous T-celllymphomas: association with disease stage and survival. Leukemia2007; 21: 2512–2518.

20 Campbell DJ, Ziegler SF. FOXP3 modifies the phenotypic andfunctional properties of regulatory T cells. Nat Rev Immunol 2007;7: 305–310.

21 Allan SE, Passerini L, Bacchetta R, Crellin N, Dai M, Orban PCet al. The role of 2 FOXP3 isoforms in the generation of humanCD4+ Tregs. J Clin Invest 2005; 115: 3276–3284.

22 Smith EL, Finney HM, Nesbitt AM, Ramsdell F, Robinson MK.Splice variants of human FOXP3 are functional inhibitors of humanCD4+ T-cell activation. Immunology 2006; 119: 203–211.

23 Lopes JE, Torgerson TR, Schubert LA, Anover SD, Ocheltree EL,Ochs HD et al. Analysis of FOXP3 reveals multiple domainsrequired for its function as a transcriptional repressor. J Immunol2006; 177: 3133–3142.

24 Bettelli E, Dastrange M, Oukka M. Foxp3 interacts with nuclearfactor of activated T cells and NF-kappa B to repress cytokine geneexpression and effector functions of T helper cells. Proc Natl AcadSci USA 2005; 102: 5138–5143.

25 Du J, Huang C, Zhou B, Ziegler SF. Isoform-specific inhibition ofROR alpha-mediated transcriptional activation by human FOXP3.J Immunol 2008; 180: 4785–4792.

26 Kaltoft K, Bisballe S, Rasmussen HF, Thestrup-Pedersen K,Thomsen K, Sterry W. A continuous T-cell line from a patientwith Sezary syndrome. Arch Dermatol Res 1987; 279: 293–298.

27 Zhang Q, Raghunath PN, Vonderheid E, Odum N, Wasik MA.Lack of phosphotyrosine phosphatase SHP-1 expression inmalignant T-cell lymphoma cells results from methylation of theSHP-1 promoter. Am J Pathol 2000; 157: 1137–1146.

28 Kaltoft K, Bisballe S, Dyrberg T, Boel E, Rasmussen PB, Thestrup-Pedersen K. Establishment of two continuous T-cell strains from asingle plaque of a patient with mycosis fungoides. In Vitro CellDev Biol 1992; 28A: 161–167.

29 Kaltoft K, Hansen BH, Pedersen CB, Pedersen S, Thestrup-Pedersen K. Common clonal chromosome aberrations in cyto-kine-dependent continuous human T-lymphocyte cell lines.Cancer Genet Cytogenet 1995; 85: 68–71.

30 Kaltoft K, Hansen BH, Thestrup-Pedersen K. Cytogenetic findingsin cell lines from cutaneous T-cell lymphoma. Dermatol Clin1994; 12: 295–304.

31 Geisler C, Scholler J, Wahi MA, Rubin B, Weiss A. Association ofthe human CD3-zeta chain with the alpha beta-T cell receptor/

FOXP3 splice forms in malignant T cellsT Krejsgaard et al

2238

Leukemia

CD3 complex. Clues from a T cell variant with a mutated T cellreceptor-alpha chain. J Immunol 1990; 145: 1761–1767.

32 Siegel JP, Mostowski HS. A bioassay for the measurement ofhuman interleukin-4. J Immunol Methods 1990; 132: 287–295.

33 Von Willebrand M, Jascur T, Bonnefoy-Berard N, Yano H,Altman A, Matsuda Y et al. Inhibition of phosphatidylinositol 3-kinase blocks T cell antigen receptor/CD3-induced activation ofthe mitogen-activated kinase Erk2. Eur J Biochem 1996; 235:828–835.

34 Sommer VH, Clemmensen OJ, Nielsen O, Wasik M, Lovato P,Brender C et al. In vivo activation of STAT3 in cutaneous T-celllymphoma. Evidence for an antiapoptotic function of STAT3.Leukemia 2004; 18: 1288–1295.

35 Woetmann A, Lovato P, Eriksen KW, Krejsgaard T, Labuda T,Zhang Q et al. Nonmalignant T cells stimulate growth of T-celllymphoma cells in the presence of bacterial toxins. Blood 2007;109: 3325–3332.

36 Zhang Q, Wang HY, Woetmann A, Raghunath PN, Odum N,Wasik MA. STAT3 induces transcription of the DNA methyltrans-ferase 1 gene (DNMT1) in malignant T lymphocytes. Blood 2006;108: 1058–1064.

37 Eriksen KW, Kaltoft K, Mikkelsen G, Nielsen M, Zhang Q, GeislerC et al. Constitutive STAT3-activation in Sezary syndrome:tyrphostin AG490 inhibits STAT3-activation, interleukin-2 receptorexpression and growth of leukemic Sezary cells. Leukemia 2001;15: 787–793.

38 Yao Z, Kanno Y, Kerenyi M, Stephens G, Durant L, Watford WTet al. Nonredundant roles for Stat5a/b in directly regulating Foxp3.Blood 2007; 109: 4368–4375.

39 Bettelli E, Carrier Y, Gao W, Korn T, Strom TB, Oukka M et al.Reciprocal developmental pathways for the generation of patho-genic effector TH17 and regulatory T cells. Nature 2006; 441:235–238.

40 Echchakir H, Bagot M, Dorothee G, Martinvalet D, Le Gouvello S,Boumsell L et al. Cutaneous T cell lymphoma reactive CD4+cytotoxic T lymphocyte clones display a Th1 cytokine profile anduse a fas-independent pathway for specific tumor cell lysis. J InvestDermatol 2000; 115: 74–80.

41 Saed G, Fivenson DP, Naidu Y, Nickoloff BJ. Mycosis fungoidesexhibits a Th1-type cell-mediated cytokine profile whereas Sezarysyndrome expresses a Th2-type profile. J Invest Dermatol 1994;103: 29–33.

42 Sakaguchi S, Sakaguchi N, Shimizu J, Yamazaki S, Sakihama T,Itoh M et al. Immunologic tolerance maintained by CD25+ CD4+regulatory T cells: their common role in controlling autoimmunity,tumor immunity, and transplantation tolerance. Immunol Rev2001; 182: 18–32.

43 Kubach J, Lutter P, Bopp T, Stoll S, Becker C, Huter E et al.Human CD4+CD25+ regulatory T cells: proteome analysisidentifies galectin-10 as a novel marker essential for their anergyand suppressive function. Blood 2007; 110: 1550–1558.

44 Paust S, Lu L, McCarty N, Cantor H. Engagement of B7 on effectorT cells by regulatory T cells prevents autoimmune disease. ProcNatl Acad Sci USA 2004; 101: 10398–10403.

45 Sors A, Jean-Louis F, Pellet C, Laroche L, Dubertret L,Courtois G et al. Down-regulating constitutive activation ofthe NF-kappaB canonical pathway overcomes the resistanceof cutaneous T-cell lymphoma to apoptosis. Blood 2006; 107:2354–2363.

46 Tobe M, Isobe Y, Tomizawa H, Nagasaki T, Takahashi H,Fukazawa T et al. Discovery of quinazolines as a novel structuralclass of potent inhibitors of NF-kappa B activation. Bioorg MedChem 2003; 11: 383–391.

47 Choi BM, Pae HO, Jeong YR, Kim YM, Chung HT. Critical role ofheme oxygenase-1 in Foxp3-mediated immune suppression.Biochem Biophys Res Commun 2005; 327: 1066–1071.

Supplementary Information accompanies the paper on the Leukemia website (http://www.nature.com/leu)

FOXP3 splice forms in malignant T cellsT Krejsgaard et al

2239

Leukemia