An activating mutation of interferon regulatory factor 4 (IRF4) in adult T-cell leukemia

The human T-cell leukemia virus-1 (HTLV-1) oncoprotein Tax drives cell proliferation and resistance to apoptosis early in the pathogenesis of adult T-cell leukemia (ATL). Subsequently, probably as a result of specific immunoediting, Tax expression is down-regulated and functionally replaced by somatic driver mutations of the host genome. Both amplification and point mutations of interferon regulatory factor 4 (IRF4) have been previously detected in ATL., K59R is the most common single-nucleotide variation of IRF4 and is found exclusively in ATL. High-throughput whole-exome sequencing revealed recurrent activating genetic alterations in the T-cell receptor, CD28, and NF-κB pathways. We found that IRF4, which is transcriptionally activated downstream of these pathways, is frequently mutated in ATL. IRF4 RNA, protein, and IRF4 transcriptional targets are uniformly elevated in HTLV-1–transformed cells and ATL cell lines, and IRF4 was bound to genomic regulatory DNA of many of these transcriptional targets in HTLV-1–transformed cell lines. We further noted that the K59R IRF4 mutant is expressed at higher levels in the nucleus than WT IRF4 and is transcriptionally more active. Expression of both WT and the K59R mutant of IRF4 from a constitutive promoter in retrovirally transduced murine bone marrow cells increased the abundance of T lymphocytes but not myeloid cells or B lymphocytes in mice. IRF4 may represent a therapeutic target in ATL because ATL cells select for a mutant of IRF4 with higher nuclear expression and transcriptional activity, and overexpression of IRF4 induces the expansion of T lymphocytes in vivo.

most commonly introduced through vertical transmission in infancy, followed by decades of asymptomatic latency. Expression of Tax, a potent oncogene and viral trans-activator, from the positive strand of the integrated HTLV-1 genome is necessary for leukemia initiation. Tax-induced activation of the NF-B pathway plays a critical role in leukemia initiation (16). Activation of other pathways, such as the PI3K-Akt-mTOR by Tax, may also play a role in transformation (17). In addition, HTLV-1 up-regulates its own transcription via Tax-induced activation of the viral 5Ј-long terminal repeat. Progression to aggressive forms of ATL is marked by decreased Tax expression, decreased transcription from the positive strand of the virus, and proviral DNA methylation or mutation, as a result of Tax-specific immunoediting (18 -21). At this stage, proliferation is sustained by cellular genes, mutated or epigenetically altered by events during leukemia initiation and progression. The identification and characterization of these cellular oncogenes and confirmation of their role as drivers of ATL will be critical for the design of new therapies.
The alterations that sustain the leukemic state recapitulate the effects of Tax, including activation of the NF-B pathway (22). Targeting NF-B is, therefore, a logical approach. However, a trial of bortezomib in ATL failed to demonstrate benefit over standard therapy (10). High-throughput whole-genome, exome, and RNA-Seq data revealed consistent overexpression and recurrent genetic alterations of the cellular proto-oncogene interferon regulatory factor 4 (IRF4/MUM1) in ATL (22). IRF4 expression is also associated with resistance to antiviral therapy (23,24). Moreover, IRF4 is expressed in ATL but not in other nonneoplastic HTLV-1-associated pathologies, such as tropical spastic paraparesis, suggesting that its role is unique to the process of malignant transformation initiated by the virus (25).
IRF4 is an established oncogene in other lymphoid malignancies, such as activated B-cell diffuse large B-cell lymphoma (ABC-DLBCL), subtypes of pediatric germinal center DLBCL (GC-DLBCL) and follicular lymphoma, Hodgkin lymphoma, and multiple myeloma. IRF4 represses key genes involved in the mitotic checkpoint, DNA repair, and apoptotic pathways (26 -31). In addition, IRF4 transforms fibroblasts in culture and is essential for Epstein-Barr virus-induced transformation of B cells in vitro, a phenomenon with many parallels to HTLV-1induced T-cell transformation (32). Therefore, IRF4 may be relevant to the design of novel therapeutic strategies for these therapy-resistant malignancies. IRF1 to -9 constitute a family of transcription factors that play key roles in innate and acquired immunity. IRF4, a lymphoid-specific IRF, was initially identified as a transcription factor that binds the 3Ј-enhancer of the immunoglobulin light chain gene as a heterodimer with the E-twenty-six (ETS) transcription factor PU.1 (Spi-1) (33,34). IRF4 has two key domains, an N-terminal DNA-binding domain and a C-terminal regulatory domain, which contains an IRF autoinhibitory domain that hydrophobically interacts with and masks the DNA-binding domain (35). Although IRF4, by itself, weakly binds DNA, binding to specific heterodimerization partners relieves autoinhibition, leading to high-affinity DNA binding (36,37). IRF4 binds as a homodimer to direct repeats of the IRF-binding element or interferon-stimulated response element (ISRE) (38).
In B cells, IRF4-PU.1 or IRF4-Spi-B heterodimers bind to ETS-IRF composite elements (EICEs) to activate transcription of B cell-specific IRF4 targets (35,39). In T cells, IRF4 complexes with BATF and JunD to bind to AP1-IRF4 composite elements (AICEs) and activate an IRF4 transcriptional program unique to T cells (40). At the intersection of IRF4 transcriptional targets and ATL-up-regulated genes lies a set of genes that are potential transcriptional targets of IRF4 in ATL and may be critical for maintaining the transformed phenotype downstream of IRF4; these genes may represent additional targets for therapy of IRF4-driven malignancies.

IRF4 is recurrently mutated in ATL
Previous studies suggested that IRF4 is recurrently mutated at K59R in ATL (Fig. 1, A and B) (22). We interrogated the COSMIC database to identify mutations of IRF4 previously detected in lymphoid malignancies. Of all single-nucleotide variations (SNVs) of IRF4 in the COSMIC database, K59R is the most common point mutation of IRF4 in lymphoid malignancies and is exclusive to ATL. We noted that other mutations of IRF4 also concentrate in the DNA-binding domain of IRF4 (Fig.  1B). We performed whole-exome sequencing on peripheral blood mononuclear cells (PBMCs) of nine patients with acute ATL. Five of nine patients (56%) had detectable somatic mutations of IRF4. Three of nine (33%) carried a K59R mutation, one patient had an L70V mutation, and one patient carried an S149N mutation. One patient, who initially presented with unfavorable chronic ATL with a white blood cell count of Ͼ50,000/l, did not carry a K59R mutation at presentation (0 of 188 reads). This patient achieved a complete remission following treatment with interferon ␣ and zidovudine but, 5 years later, relapsed with lymphoma subtype of ATL with bulky lymphadenopathy. Repeat sequencing at that time revealed the IRF4 K59R mutation in 30 of 156 reads, suggesting that the emergence of this mutation correlated with progression of disease.
Lysine 59 of IRF4 lies within the DNA-binding domain of IRF4 (Fig. 1B), which has a helix-loop-helix tertiary structure with five conserved tryptophans, which are characteristic of the DNA-binding domains of interferon regulatory factors (41). Comparison of the amino acid sequence adjacent to the mutated lysine in IRF4 reveals that this region is highly conserved across evolutionarily diverse species (Fig. 1C), suggesting that this is a functionally important region of the protein. Moreover, the Lys-59 position in IRF4 corresponds to either a lysine or arginine in all other IRFs (Fig. 1D), suggesting that the K59R mutation is unlikely to result in loss of function.
IRF4 lies downstream of mutations in the TCR, CD28, and NF-B signaling pathways, suggesting that it is a rational target for further study and for the design of future ATL therapies. Kataoka et al. (22) demonstrated that these pathways are frequently activated in ATL, as a result of gene amplification or gain-of-function mutations in Vav1, phospholipase C␥1, pro-

IRF4 is highly expressed in HTLV-1-transformed and ATL cell lines
To determine the level of expression of IRF4 protein in HTLV-1-transformed and ATL cell lines (Fig. 3A), we per-

IRF4 is bound to genomic regulatory DNA of transcriptional targets
To determine whether IRF4 is bound to its transcriptional targets in HTLV-1-transformed and ATL cell lines, we performed ChIP assays on CD3/CD28-activated PBMCs and Jur-kat, MT2, MT4, TL-OM1, and MT1 cells with an IRF4-specific goat polyclonal antibody and control goat immunoglobulin (Fig. 3B). We measured binding of IRF4 to the 3Ј-regulatory region of the HELIOS gene. Levels of IRF4 bound to this gene in PBMCs were 0.47% of input compared with 0.01% background levels as measured by qPCR. Jurkat cells showed no significant binding of IRF4. In contrast, IRF4 binding to the HELIOS gene was 2.2-, 9.7-, 330-, and 15.5-fold greater than background levels in MT2, MT4, TL-OM1, and MT-1 cells, respectively. There was a statistically significant difference of HELIOS DNA content between control and IRF4 immunoprecipitates from PBMCs and MT4, TL-OM1, and MT1 cells.
CTLA4 is an immune checkpoint receptor and transcriptional target of IRF4 that is overexpressed in ATL (40). We assayed for the presence of CTLA4 DNA in control IgG and IRF4 immunoprecipitates (Fig. 3C). Activated PBMC control immunoprecipitates and IRF4 immunoprecipitates yielded 0.03 and 0.42% of input DNA, respectively, as measured by   (22). SNVs are depicted in black type and copy number variations (CNV) in white type. Note that in ATL, protein kinase C␤ is exclusively altered by point mutation and protein kinase C exclusively by amplification.

IRF4 activation in ATL
qPCR. Jurkat cells showed no significant binding of IRF4 to the CTLA4 gene. In contrast, levels of IRF4 binding to the CTLA4 gene were 1.7-, 5.1-, 9.6-, and 7.8-fold above background in MT2, MT4, TL-OM1, and MT1 cells, respectively. A statistically significant difference of CTLA4 DNA content between control and IRF4 immunoprecipitates was detected in PBMCs, MT4, and MT1 cells.
As a control, we assayed for a control region 10 kilobases upstream of the interferon ␤ start site in control and IRF4 chromatin immunoprecipitates by qPCR. There was no significant difference in levels of this DNA region between control and IRF4 chromatin immunoprecipitates of any of the assayed T-cell lines.

IRF4 transcript is highly expressed in HTLV-transformed and ATL cell lines
We measured IRF4 transcript in a control HTLV-1-negative T-ALL cell line (Jurkat) and HTLV-1-positive cell lines MT2, MT4, TL-OM1, and MT1 (Fig. 4A). As compared with Jurkat cells, 37-, 121-, 10-, and 25-fold levels of expression of IRF4 transcript were observed in MT2, MT4, TL-OM1, and MT1 cells, respectively. These results were all statistically significant. We sequenced the entire DNA-binding domain coding sequence of IRF4 in MT2, MT4, TL-OM1, ED40515, and MT1 cells; we did not detect any mutations in the DNA-binding domain coding sequence of any of these cell lines.

Potential functional consequences of the K59R mutation
Whereas mutation of Lys-59 of IRF4 is unique to ATL, the corresponding residue in other IRF family members is frequently altered in malignant cells. Mutations in IRF2 Arg-43 and IRF6 Arg-45 have been described in cervical and ovarian cancers. Interestingly, IRF8 K45R is found in aggressive forms of DLBCL (65). The mutation K59R is a conservative change that replaces a positively charged lysine with arginine, which possesses an isoelectric pH that is higher than that of lysine (10.7 versus 9.7), and therefore it is likely to increase the net positive charge in this region. We therefore hypothesized that this amino acid change might preserve DNA binding while preventing ubiquitination, thus leading to increased protein levels or enhanced DNA binding and increased transcriptional activity. It is interesting to note that a K123R mutation of IRF4 detected in 2% of patients with chronic lymphocytic leukemia and multiple myeloma results in increased protein expression of the mutant protein as compared with the WT protein (66,67). Alternatively, given the concentration of positively charged residues in this region of IRF4, K59R could lead to changes in subcellular localization, although we were unable to identify a canonical nuclear localization or nuclear export signal in this region.

IRF4 activation in ATL
blots demonstrate purity of the nuclear and cytoplasmic fractions, respectively (middle and lower blots of Fig. 5, A and B).

K59R mutation enhances nuclear expression levels of IRF4 protein without affecting cytosolic levels
To assess the effects of the K59R mutation on subcellular localization of IRF4, 293T cells were transfected with equal amounts of either empty murine stem cell virus (MSCV)-IRES-GFP vector or corresponding WT or K59R IRF4 expression vectors. Forty-eight hours after transfection, subcellular fractionation was performed, and nuclear and cytosolic fractions were separated by SDS-PAGE and then analyzed by immunoblot (Fig. 5C). Lanes 1, 2, and 3 show nuclear expression of IRF4 and histone H3 (nuclear loading control) and lack of nuclear expression of HSP90, which is a cytosolic protein. IRF4 was detected with an antibody to the C terminus of IRF4, a region that is least conserved among different IRFs, which is distal to the Lys-59 site and unlikely to affect antibody binding. This experiment was confirmed with three biological replicates and with two different antibodies to distinct epitopes at the C terminus of IRF4. Densitometry of IRF4 blots revealed a statistically significant 2.8-fold increase in nuclear levels of K59R mutant IRF4 as compared with WT IRF4 (Fig. 5C, p ϭ 0.03).
To assess whether similar findings are found in T-cell lines, we transfected, by electroporation, Jurkat T cells with 50 g of either MSCV-Empty vector, MSCV-IRF4 WT, or MSCV-IRF4 K59R mutant plasmid and analyzed the localization of IRF4 by subcellular fractionation of cells at 24 h post-transfection (Fig.  5D). The IRF4 blot (first blot from the top) shows 6.2-fold increased nuclear levels of IRF4 in K59R mutant-transfected Jurkat cells as compared with WT IRF4 -transfected cells. Cytoplasmic levels of IRF4 were not different. Note that there is a nonspecific band in the empty vector-transfected lane of the cytoplasmic fraction (lane 4, first blot from the top). Whole-cell extracts showed 2.7-fold increased levels of K59R mutant IRF4 as compared with WT IRF4, which is accounted for entirely by the increased nuclear levels of the K59R mutant. A histone H3 blot (second from the top) and HSP90 blot (third from the top) indicate the purity of the subcellular fractions.

K59R mutant IRF4 possesses increased transcriptional activity as compared with WT IRF4
To assess the transcriptional activity of the mutant compared with WT IRF4, we used luciferase reporter plasmids. We co-

IRF4 activation in ATL
transfected 293T cells with interferon ␤ firefly luciferase (interferon ␤ Luc) plasmid and TK-Renilla (Renilla luciferase with the thymidine kinase promoter) and either MSCV-IRES-GFP empty vector or increasing quantities of vectors expressing WT or K59R IRF4 (Fig. 6A). Empty vector and IRF4 WT plasmids were transfected in a 3:1 ratio to K59R IRF4 plasmids to equalize nuclear levels based on our prior ϳ3:1 K59R/WT IRF4 nuclear expression ratios. The difference in mass of transfected DNA was compensated by pcDNA3 plasmid, so that total transfected DNA remained constant. We observed 3-, 4-, and 5-fold induction of luciferase with MSCV-IRES-IRF4 WT relative to empty vector with 0.75, 1.5, and 3 g of plasmid, respectively. With one-third the mass of K59R IRF4 expression plasmid (0.25, 0.5, and 1 g) we observed 3.7-, 8.4-, and 11.6-fold induction of luciferase, respectively. The differences in results between 1.5 g of WT IRF4 plasmid and 0.5 g of K59R mutant IRF4 expression plasmid and between 3 g of WT IRF4 expression plasmid and 1 g of K59R expression plasmid were statistically significant (five replicates). We ensured that the nuclear expression level of K59R did not exceed that of WT IRF4 by separating nuclear extracts by SDS-PAGE and immunoblotting for IRF4 (Fig. 6B). The IRF4 immunoblot shows the amount of nuclear IRF4 protein increasing with increasing amounts of transfected plasmid for WT IRF4 (lanes 2-4) and for K59R mutant (lanes 5-7). The largest difference in luciferase activation was between 3 g of transfected WT IRF4 plasmid and 1 g of K59R mutant plasmid (bar 4 versus bar 7; 5versus 11.6fold induction); the corresponding lanes in the IRF4 immunoblot showed increased expression of IRF4 WT (lane 4) compared with K59R mutant IRF4 (lane 7), ruling out increased nuclear levels as the cause of this difference.

Transcriptional activity of WT and K59R mutant IRF4 at AICEs, EICEs, and ISREs
IRF4 heterodimerizes with BATF and JunD to activate transcription from AICEs (40) and with Ets family transcription factors PU.1 and SpiB to activate transcription from EICEs (68). IRF4 also homodimerizes to activate direct repeats of the IRF binding sequence (GAAA), also known as an ISRE (34). We tested the capacity of WT IRF4 and K59R mutant IRF4 to induce transcription from these binding elements. For these experiments, we utilized S-tagged IRF4 constructs cloned into the TriExNeo backbone because these plasmids expressed protein at significantly higher levels than the MSCV vectors. We found consistent induction of the interferon ␤ Luc reporter with S-tagged WT IRF4 (10-fold) and K59R IRF4 (23-fold), demonstrating that the transcriptional potential of the S-tagged constructs is similar to that of the untagged protein (Fig. 7A). AICE-Luc, incorporating three repeats of the AICE sequence, was co-transfected with TK-Renilla and either empty S-tagged vector, S-tagged IRF4 WT, or S-tagged K59R mutant IRF4 expression plasmids (Fig. 7B). We found 1.3-fold activation with WT IRF4 and 1.6-fold activation with K59R IRF4 mutant (four replicates). There was a statistically significant difference between induction with K59R mutant and WT IRF4; however, induction of the promoter by IRF4 was weak.
We next assayed induction of the AICE Luc construct with WT and K59R mutant IRF4 in combination with BATF (Fig.  7C). We found 1.4-and 1.6-fold induction of AICE Luc with WT and K59R mutant IRF4, respectively (four replicates). The difference between empty vector and K59R IRF4 was statistically significant, although there was no statistically significant difference between the induction of WT and K59R IRF4.
We then assayed the induction of EICE-Luc by WT IRF4 and K59R IRF4. EICE-Luc incorporating three repeats of the EICE sequence was co-transfected with TK-Renilla and either empty vector or S-tagged IRF4 WT or S-tagged K59R mutant IRF4 expression plasmids (Fig. 7D). We observed 1.0-and 1.2-fold induction of EICE-Luc with WT IRF4 and K59R IRF4, respectively (four replicates). The differences between K59R and empty vector and K59R and WT IRF4 were statistically significant.
We assayed induction of EICE Luc by WT IRF4 and K59R mutant IRF4 in combination with the prototypical Ets family transcription factor PU.1 (Fig. 7E). We observed 0.9-and 1.2fold induction with WT IRF4 and K59R IRF4, respectively (four replicates). These values were not statistically significant.
We next assayed the induction of ISRE Luc by WT IRF4 and K59R IRF4. ISRE-Luc incorporating three repeats of ISRE sequence was co-transfected with TK-Renilla and either empty vector or S-tagged IRF4 WT or S-tagged K59R mutant IRF4

IRF4 activation in ATL
expression plasmids (Fig. 7F). We observed 2.4-and 2.9-fold induction of ISRE Luc with co-transfection WT IRF4 and K59R mutant IRF4, respectively (four replicates). There was a statistically significant difference in luciferase values between empty vector and WT IRF4 and between empty vector and K59R mutant IRF4, but not between WT IRF4 and K59R IRF4.
Given that we observed only weak induction of AICE Luc and EICE Luc even with co-transcriptional activators of these elements, and weak induction of ISRE Luc by IRF4, we further examined the interferon ␤ promoter, which was strongly induced by IRF4. Examination of this sequence revealed mirror repeats of the IRF-binding element (TTTC-CTTT/AAAGGAAA), which is distinct from previously described AICEs, EICEs, and ISREs. We hypothesized that this sequence might account for the strong induction of the interferon ␤ promoter by IRF4. We cloned three repeats of this sequence upstream of the luciferase gene and tested this luciferase reporter construct for induction by IRF4 (Fig. 7H). We observed 39-and 85-fold induction of luciferase with IRF4 WT and K59R IRF4, respectively. The differences between WT IRF4 and empty vector, between K59R IRF4 and empty vector, and between WT IRF4 and K59R IRF4 were statistically significant. This would suggest that this mirror repeat sequence, which allows for strong induction by IRF4, is a binding site for IRF4 in the interferon ␤ promoter.

IRF4 knockdown increases apoptosis of an ATL cell line
Given the finding of recurrent mutations of IRF4 in ATL, the increased nuclear localization of the mutant, and increased transcriptional activity of the mutant, we further examined the effects of IRF4 on the viability of an ATL cell line. HuT102 cells were electroporated with control siRNA or an IRF4-targeted pool of siRNAs; cells were lysed at 72 h, and lysates were sub-jected to analysis by immunoblot (Fig. 8A).

IRF4 increases abundance of murine T cells
In light of the finding of increased apoptosis in IRF4 knockdown cells in vitro, we further examined the effects of IRF4 on T-cell abundance in vivo. C57/B6 WT mice were transplanted with bone marrow-derived CD45.1 ϩ Sca-1 ϩ cells, which were transduced MSCV-IRES-GFP empty retroviral vectors or IRF4 WT-or IRF4 K59R-expressing vectors (Fig. 8, B and C). Four weeks after engraftment, peripheral blood was used for FACS analysis of donor cells. GFP ϩ CD45.1 ϩ cells (donor cells successfully transduced with the MSCV-IRES-GFP retroviral vectors) were classified as myeloid based on reactivity for Gr-1 and Mac-1; B lymphocytes based on reactivity for B220; and T lymphocytes based on reactivity for CD4 or CD8. Fifteen replicates were performed with control mice and 10 replicates each of WT IRF4 -and K59R-expressing mice (Fig. 8, A and B). We observed statistically and biologically significant increases in GFP ϩ donor T cells, but not for B lymphocytes or myeloid cells in mice that were transplanted with Sca-1 ϩ bone marrow stem cells transduced with WT IRF4 and K59R mutant IRF4. This A, -fold induction of firefly luciferase expressed from the interferon ␤ proximal promoter by WT S-tagged IRF4 and S-tagged K59R mutant IRF4 over empty vector expressed as a ratio to Renilla luciferase expressed from thymidine kinase promoter. B, induction of firefly luciferase expressed from an AP1-IRF composite element by S-tagged WT IRF4 and S-tagged K59R mutant IRF4 expressed as a ratio to Renilla luciferase expressed from thymidine kinase promoter. C, induction of firefly luciferase expressed from an AP1-IRF composite element by S-tagged WT IRF4 and S-tagged K59R mutant IRF4 transfected in combination with BATF expressed as a ratio to Renilla luciferase expressed from thymidine kinase promoter. D, induction of firefly luciferase expressed from an Ets-IRF composite element by S-tagged WT IRF4 and S-tagged K59R mutant IRF4 expressed as a ratio to Renilla luciferase expressed from thymidine kinase promoter. E, induction of firefly luciferase expressed from an Ets-IRF composite element by S-tagged WT IRF4 and S-tagged K59R mutant IRF4 in combination with PU.1 expressed as a ratio to Renilla luciferase expressed from thymidine kinase promoter. F, induction of firefly luciferase expressed from an ISRE element by WT IRF4 and K59R mutant IRF4 expressed as a ratio to Renilla luciferase expressed from thymidine kinase promoter. G, sequence from the IFN-␤ Luc plasmid corresponding to nucleotides Ϫ259 to ϩ43 relative to the transcription start site was sequenced from the interferon ␤ Luc plasmid used in experiments depicted in Figs. 6A and 7A. A mirror repeat element in that sequence is highlighted. H, induction of firefly luciferase expressed from a mirror repeat element by S-tagged WT IRF4 and S-tagged K59R mutant IRF4 expressed as a ratio to Renilla luciferase expressed from thymidine kinase promoter. A-G, results of four replicates; H, results of three replicates.

IRF4 activation in ATL
demonstrates that IRF4 selectively increased T-lymphocyte abundance in vivo.

Discussion
IRF4 is a key oncogenic transcription factor in common lymphoid malignancies, such as ABC-DLBCL and multiple myeloma, which are characterized by the activation of NF-B (27,28,67,68). HTLV-1-associated malignancies are initially sustained by the potent viral oncogene Tax. Subsequently, Tax expression is down-regulated to evade Tax-specific immunoediting, and somatically mutated host proteins take over as drivers of proliferation and confer resistance to apoptosis and therapy. ATL is characterized by mutations in the TCR, CD28, and NF-B signaling pathways. IRF4 is transcriptionally upregulated by all of these pathways. Moreover, IRF4 is genetically altered by amplification in 25% of ATL patients and point mutation in 14% of patients (22). Mutations at K59R and L70V are recurring mutations in ATL and represent the most common mutations of IRF4 identified in the COSMIC database of cancer-associated mutations (22). Our exome sequencing data revealed that 33% of acute ATL patients had a K59R mutation, and 11% of patients had an L70V mutation. In one patient, the K59R mutation emerged at progression from chronic ATL to acute ATL, suggesting that this mutation arose with disease progression. Therefore, there was a strong rationale for further characterization of the role of IRF4 and the functional consequences of IRF4 mutations in lymphoid malignancies. We did not detect mutations of the IRF4 DNA-binding domain in the HTLV-1-transformed cell lines we studied. It is possible that selection of this mutation only occurs in vivo; however, the small number of available cell lines precludes this conclusion.
We observed significantly increased IRF4 transcript and protein expression in HTLV-1-positive cell lines as compared with HTLV-1-negative cell lines and activated PBMCs, consistent with a role for IRF4 in their continued viability. We also observed greater binding of IRF4 to regulatory genomic DNA of its transcriptional targets in HTLV-1-positive cell lines and CD3/CD28 activated PBMCs as compared with control Jurkat cells.
We measured levels of transcriptional targets of IRF4, including BATF, CTLA4, CCR4, cFLIP, IL-2, IL-9, and IL-10, which were transcriptionally up-regulated, and also Helios transcripts, which were down-regulated by IRF4. We did not see a perfect correlation between IRF4 protein levels and expression levels of its transcriptional targets in HTLV-transformed and ATL cell lines, suggesting that other transcription factors may also be involved in their regulation.
We noted that the K59R mutation of IRF4 is in a highly conserved region of the DNA-binding domain. We hypothesized that this might affect protein levels, subcellular localization, or transcriptional activation potency. We noted that IRF4 is predominantly nuclear in HTLV-1-transformed and ATL cell lines. We observed increased nuclear expression of K59R mutant IRF4 as compared with WT protein, without changes in cytosolic levels in transiently transfected 293T cells and Jurkat T cells. This suggests that the difference in nuclear levels may be due to differences in import or to sequestration by other nuclear localized molecules.

Figure 8. Effect of WT and K59R IRF4 on T-lymphocyte abundance in mice.
A, Western blotting depicting expression of cleaved caspase-3 (as a measure of apoptosis), total caspase-3, IRF4, Tax, and actin in lysates of HuT102 cells treated with control siRNA or a pool of siRNAs to IRF4. B and C, murine hematopoietic stem cells (CD45.1, c-kit ϩ , Sca-1 ϩ bone marrow cells) were transduced with MSCV retroviral vectors, which express IRF4, IRF4 (K59R), or vector alone, as well as GFP from an IRES and engrafted into irradiated CD45.2 recipients (n ϭ 15, 10, and 10 for control, WT IRF4 -overexpressing, and K59R-overexpressing mice). Four weeks after engraftment, peripheral blood was collected, and FACS was used to determine the abundance of CD45.1, GFP ϩ myeloid cells (Gr-1ϩ, Mac-1ϩ), B cells (B220 ϩ ), and T cells (CD4 ϩ , CD8 ϩ ). Whereas no effect was observed on B cells or myeloid cells, the T-cell populations in IRF4-overexpressing mice were significantly greater than controls, demonstrating that overexpression of IRF4 or IRF4 (K59R) results in T-cell proliferation in vivo. D, mechanisms for increased oncogenic activity of the ATL-specific IRF4 K59R mutation.

IRF4 activation in ATL
When we tested the capacity of K59R mutant IRF4 to activate the interferon ␤ promoter, we observed that at equal nuclear expression levels, the K59R mutant possessed more than twice the transcriptional activation potency as compared with WT IRF4 . Surprisingly, when we tested the induction of more conventional IRF4-responsive promoter elements (i.e. AP1-IRF and Ets-IRF composite elements and interferon-stimulated response elements), we saw less increase over background. Although these promoters were also activated more potently by K59R mutant IRF4 than WT IRF4, the differences between signal and background were smaller.
Given the strong induction of the interferon ␤ promoter by IRF4, we examined the promoter sequence for IRF-binding sequences. We observed a mirror repeat of the IRF response element in the proximal promoter. When we tested the effect of IRF4 on this promoter element, we saw strong induction, which would fully account for the high level of induction of the interferon ␤ promoter by IRF4. When we measured binding of IRF4 to the interferon ␤ promoter in HTLV-positive cell lines, CD3/ CD28-activated PBMCs, and Jurkat cells by a ChIP assay, we observed binding in TL-OM1 cells and MT1 cells, both of which are ATL-derived cell lines. Although interferon ␤ activates apoptosis of naive T cells, it allows for clonal expansion of TCR-activated T cells in a cell-intrinsic manner and can inhibit activation-induced cell death (69 -73). However, it is unclear whether interferon ␤ plays a role in survival of ATL cells. It is also possible that other genes with the same mirror repeat IRFbinding element may be the key targets of IRF4 required for viability of ATL cells.
When we knocked down IRF4 expression in HuT102 ATL cells in vitro, we found increased apoptosis, suggesting that ATL cell lines depend on IRF4 expression for viability. When we tested the capacity of hematopoietic cells transduced with IRF4 to expand in vivo, we noted a selective effect of IRF4 on T cells, further supporting a role for IRF4 in T-cell malignancies. We did not see a statistically significant difference in T-cell numbers between mice that received WT IRF4 -transduced syngeneic bone marrow cells and those that received K59R mutant IRF4 -transduced cells. It is unclear why a difference was not identified, but it is possible that the increased activity of K59R mutant IRF4 compared with WT IRF4 is only manifested in human cells due to interactions with co-transcriptional activators.
These data suggest that novel IRF4-targeted agents may have efficacy in ATL and other IRF4-driven T-cell malignancies. The immunomodulatory agent lenalidomide had activity in a phase II trial in relapsed/refractory ATL (11). Lenalidomide has been shown to depress IRF4 levels in multiple myeloma cell lines, and sensitivity to lenalidomide correlates with IRF4 levels (74). We did not see inhibition of ATL cell lines by lenalidomide in vitro (data not shown), suggesting that, at least in ATL, responses to lenalidomide may not be due to cell-intrinsic effects. IRF4 is critical for the development of conventional type 2 dendritic cells, which drive Th2, Th9, and Th17 responses, as opposed to type 1 conventional dendritic cells, which drive Th1 responses (75). It is possible that in ATL, lenalidomide may skew immune responses toward a Th1-biased response, at the cost of a Th2 response to tumor-associated antigens, as demonstrated in multiple myeloma (76,77).
In summary, IRF4 is highly expressed in ATL cell lines and HTLV-1-transformed cell lines and stimulates the proliferation of T lymphocytes. Moreover, the K59R mutant form of IRF4 is expressed at higher levels in the nucleus and possesses increased transcriptional activity compared with WT IRF4, strongly suggesting that this is an activating mutation in ATL.

Experimental procedures
Cell culture TL-OM1, MT1, and ED40515 Tax-negative ATL cell lines were obtained from Edward Harhaj (Johns Hopkins University). The HuT102 ATL cell line was a gift from Patrick Green (Ohio State University). Jurkat, 293T, MT2, and MT4 cells were obtained from the National Institutes of Health AIDS Repository.
Cell lines were maintained at 37°C and 5% CO 2 in complete medium supplemented with 10% fetal bovine serum, 4 mM L-glutamine, 100 units/ml penicillin, 100 g/ml streptomycin, and 0.25 g/ml amphotericin B. PBMCs and T-cell lines, including Jurkat, MT2, MT4, TL-OM1, MT1, and ED40515 cells, were maintained in complete RPMI medium. 293T cells were maintained in complete Dulbecco's modified Eagle's medium with 1 mM sodium pyruvate. For CD3/CD28 stimulation of PBMCs, preservative-free anti-CD3 (clone OKT3) and anti-CD28 (clone ANC28) at 1 g/ml in PBS were bound to tissue culture-treated polystyrene plates by incubation overnight. Plates were then blocked for 30 min with complete RPMI medium before application of cells.

Transfections
293T cells were transiently transfected with Transit (Mirus Bio) as per the manufacturer's recommended protocol. Jurkat cells were transfected with 50 g of MSCV-Empty vector, MSCV-IRF4 WT, or MSCV-IRF4 K59R by electroporation at 340 V with 400-microfarad capacitance and parallel resistance of 124 ohms in a volume of 150 l of electroporation buffer consisting of RPMI without antibiotics, bicarbonate, or glutamine with 10 mM glycylglycine, pH 8.3, in a 4-mm gap cuvette (Bulldog Bio). HuT102 was transfected with either 75 pmol of control siRNAs or an siRNA pool targeting the IRF4 3Ј-UTR by electroporation at 325 V with 400-microfarad capacitance and parallel resistance of 124 ohms in a volume of 150 l.

Plasmids
Plasmids expressing WT IRF4 and K59R mutant IRF4 were constructed by recombinational cloning of the respective cDNAs using LR clonase into MSCV gateway vector. S-tagged IRF4 and K59R mutant IRF4 constructs were made by amplifying the IRF4 coding sequences from the MSCV vectors described above using the left primer CGCGGATCCGATGA-ACCTGGAGGGCGG and right primer CGCGAATTCTCAT-TCTTGAATAGAGGAA, which introduced BamHI and EcoRI sites in termini of the amplified sequences that were ligated into IRF4 activation in ATL the corresponding restriction sites in the Tri Ex Neo backbone vector, so that WT IRF4 and K59R mutant IRF4 were expressed with a hexahistidine tag and an S tag at the N terminus.
Interferon ␤ Luc plasmid incorporating two repeats of nucleotides Ϫ259 to ϩ43 of the interferon ␤ proximal promoter was a gift from the late Dr. Paula Marie Pitha-Rowe (Johns Hopkins University). For the construction of AICE-Luc, EICE-Luc, ISRE-Luc, and (TTTCCTTT) 3 firefly luciferase reporter plasmids, pairs of partially complementary 5Ј-phosphorylated oligonucleotides incorporating three repeats of AICE, EICE, ISRE, and TTTCCTTT mirror repeat sequences were annealed and ligated with 5Ј NheI and 3Ј HindIII sites in PGL4.25Luc2CPminP plasmid, obtained from Promega Corp.

Subcellular fractionation
Cells were washed twice with ice-cold PBS and resuspended in hypotonic buffer (20 mM Tris-HCl, pH 7.4, 10 mM NaCl, and 3 mM MgCl 2 ). One-twentieth volume of 10% Igepal CA630 was added to each sample, and samples were vortexed vigorously before spinning down nuclei at 8,000 rpm at 4°C in a refrigerated microcentrifuge. Supernatants were used as cytosolic fractions. Nuclear pellets were lysed in Tris-HCl, pH 6.8, with 2% Igepal CA630, 2% SDS, 10 mM sodium ␤-glycerophosphate, 10 mM sodium fluoride, 2.5 mM sodium pyrophosphate, 1 mM sodium orthovanadate, and EDTA-free protease inhibitor mixture (Roche Applied Science) and sonicated on ice for 20 s before use for analysis (modified from Ref. 78).

PAGE and Western blotting
Cells were lysed in Tris-HCl, pH 6.8, with 2% Igepal CA630, 2% SDS, 10 mM sodium ␤-glycerophosphate, 10 mM sodium fluoride, 2.5 mM sodium pyrophosphate, 1 mM sodium orthovanadate, and EDTA-free protease inhibitor mixture (Roche Applied Science/Sigma). Lysates were sonicated on ice for 20 s. Protein concentration was determined using the bicinchoninic acid assay (79). Equal amounts of protein (40 g) were loaded onto polyacrylamide gels, and electrophoresis was performed according to methods described by Ornstein (80) and modified by Laemmli (81). Primary antibodies were diluted in 5% protease-free BSA, and secondary antibodies were diluted in 10% skim milk. Blots were imaged by enhanced chemiluminescence using a Chemidoc imager (Bio-Rad) and quantified using the Bio-Rad proprietary software.
The following antibodies were used for Western blots. Rabbit primary polyclonal antibody to HSP90 (SPA46) was obtained from Stressgen Corp.; rabbit primary mAb to histone H3 clone D1H2 was obtained from Cell Signaling Technology (catalog no. 4499); D175 rabbit primary polyclonal antibody to cleaved caspase-3 was obtained from Cell Signaling Technology (catalog no. 9661); mouse mAb to caspase-3 clone 3G2 was obtained from Cell Signaling Technology (catalog no. 9668); goat polyclonal antibody to IRF4 (M17; sc6059) was obtained from Santa Cruz Biotechnology, Inc.; goat polyclonal antibody to IRF4 was obtained from Sigma-Aldrich (SAB2502031); rabbit polyclonal antibody to full-length Aequorea coerulescens GFP was obtained from Clontech (catalog no. 632459); hybridoma 1316 for Tax1 was obtained from the NIH AIDS Repository.
ChIP T-cell lines and CD3/CD28-activated (48 h) PBMCs were washed once in ice-cold PBS and resuspended in 2% paraformaldehyde in PBS for 5 min for cross-linking at room temperature. The reaction was quenched by resuspension in 200 mM glycine for 10 min. Cells were washed in PBS and then lysed in Tris-HCl, pH 7.4, 150 milliequivalents/liter NaCl, 0.1% sodium deoxycholate, 200 M putrescine, and 10% glycerol. The lysate was sonicated in a bath sonicator with controlled temperature Ͻ10°C for 20 min. The lysates were cleared by a brief centrifugation at 15,000 rpm for 10 min in a refrigerated centrifuge. Supernatants were used for ChIP analysis. DNA in each sample was quantified using Hoechst 33342 fluorescence, and equal mass of DNA from each sample was included in the immunoprecipitation reaction. Goat anti-IRF4 polyclonal antibody (M-17 from Santa Cruz Biotechnology) and magnetic protein G beads (Thermo Fisher Scientific) were added to the immunoprecipitation reaction overnight. After five washes with Tris-HCl, pH 7.4, 150 milliequivalents/liter NaCl, and 200 M putrescine, protein and DNA were eluted from the beads by incubation at 65°C overnight, which also served to reverse DNA-DNA and DNA-protein cross-links. Samples were simultaneously treated with RNase A overnight. Samples were used for qPCR to quantify genomic areas of interest in the IRF4 immunoprecipitates. Immunoprecipitates with equal amounts of DNA with IgG control were performed in parallel to determine the background signal by qPCR.
Genomic DNA was used as template for amplification in a three-step amplification protocol using iTaq Universal SYBR Green one-step kit on a Bio-Rad CFX Connect real-time system with an extension temperature of 65°C. Primers were designed

IRF4 activation in ATL
to flank IRF4-binding sites in the genomic neighborhood of genes of interest, as determined using ENCODE data viewed with the Integrated Genome Viewer (Broad Institute).
Input DNA was diluted 100-fold before qPCR. Cycle number (Ct) for immunoprecipitated DNA was converted to percentage of input by dividing antilog to the base 2 of (Ϫ1 ϫ Ct) for immunoprecipitates by antilog to the base 2 of (Ϫ1 ϫ Ct) for input to give a percentage of input (accounting for the dilution factor of 100 for input DNA).

Genomic DNA sequencing
Genomic DNA from HTLV-1-transformed cell lines was used as template to amplify sequence corresponding to exons 2 and 3 of the IRF4 gene. Forward and reverse primers used for PCR were used for sequencing reactions to obtain forward and reverse sequences. Sequences for forward and reverse primers for amplification of exon 2 with the 3Ј position from the IRF4 transcription start site indicated are as follows: IRF4 -1351 (forward), CCCCGCAGTGCAGAGCAGA; IRF4 -1677 (reverse), GGGGACCCCGGGCTCTGTCT.

Whole-exome sequencing
Genomic DNA was extracted from PBMCs of nine ATL patients and sonicated to generate 200 -300-base pair fragments. Exonic DNA fragments were captured with biotin-labeled probes covering the entire human exome. Pulldown was performed with streptavidin beads, and eluted DNA was used to make libraries with Illumina linkers and amplified using specific index primers for each patient's DNA sample. For the sequencing process, the ultra-high-throughput sequencing system, HiSeq 2500, was used, generating paired-end reads of 250 nucleotides each. Alignment and variant calling was performed using Basespace.

qRT-PCR
RNA was extracted using TRIzol reagent. Total RNA was used as template for reverse transcription and amplification of cDNAs in a three-step amplification protocol using the iTaq Universal SYBR Green one-step kit (Bio-Rad) on a Bio-Rad CFX Connect real-time system with an extension temperature of 65°C. Exon-exon junction-spanning primers were designed to reduce background from genomic DNA (except for interferon ␤, which has no introns). The following primers were used for qRT-PCR: HGPRT1 ( Cycle number (Ct value) for each measured transcript was converted to ratios of transcript abundance by taking the antilog to the base 2 of (Ϫ1 ϫ Ct) value for each transcript and dividing that by the antilog to the base 2 of (Ϫ1 ϫ Ct) value for hypoxanthine guanine phosphoribosyl transferase based on the assumption of a 100% amplification efficiency.

Transduction of murine hematopoietic stem cells and transplantation into recipient mice
Donor B6 CD45.1 mice were treated with 150 mg/kg 5-fluorouracil 6 days before marrow harvest. Harvested bone marrow was enriched for Sca-1 ϩ cells by autoMACS (Miltenyi). Cell concentrations were adjusted to 5 ϫ 10 5 cells/ml in medium containing Stempro34 (Gibco) serum-free medium supplement, 100 units/ml penicillin, 100 g/ml streptomycin, 2 mM glutamine, 10 ng/ml stem cell factor (R&D Systems), 100 ng/ml mouse thrombopoietin, and 4 g/ml Polybrene. Virus-like particles were generated in 293T cells by calcium phosphatemediated co-transfection of pCL-ECO packaging plasmid with either the empty MSCV-IRES-GFP retroviral vector or vector with coding sequence for WT IRF4 or K59R mutant IRF4 inserted by LR clonase recombinational cloning. Virus-like particles were harvested from the medium by ultracentrifugation. Sca-1 ϩ bone marrow cells were transduced with MSCV retroviral vectors by spin-infection at 250 ϫ g at room temperature for 2 h, followed by incubation for 1 h at 37°C, and then transplanted by retro-orbital injection into lethally irradiated (9.5 grays) CD45.2 (C57/B6 WT) recipients (n ϭ 15 for empty vector and 10 each for WT IRF4 and mutant IRF4). Four weeks after engraftment, peripheral blood was collected, and FACS was used to determine the abundance of CD45.1 ϩ , GFP ϩ mye-