Super-enhancer-based identification of a BATF3/IL-2R−module reveals vulnerabilities in anaplastic large cell lymphoma

Anaplastic large cell lymphoma (ALCL), an aggressive CD30-positive T-cell lymphoma, comprises systemic anaplastic lymphoma kinase (ALK)-positive, and ALK-negative, primary cutaneous and breast implant-associated ALCL. Prognosis of some ALCL subgroups is still unsatisfactory, and already in second line effective treatment options are lacking. To identify genes defining ALCL cell state and dependencies, we here characterize super-enhancer regions by genome-wide H3K27ac ChIP-seq. In addition to known ALCL key regulators, the AP-1-member BATF3 and IL-2 receptor (IL2R)-components are among the top hits. Specific and high-level IL2R expression in ALCL correlates with BATF3 expression. Confirming a regulatory link, IL-2R-expression decreases following BATF3 knockout, and BATF3 is recruited to IL2R regulatory regions. Functionally, IL-2, IL-15 and Neo-2/15, a hyper-stable IL-2/IL-15 mimic, accelerate ALCL growth and activate STAT1, STAT5 and ERK1/2. In line, strong IL-2Rα-expression in ALCL patients is linked to more aggressive clinical presentation. Finally, an IL-2Rα-targeting antibody-drug conjugate efficiently kills ALCL cells in vitro and in vivo. Our results highlight the importance of the BATF3/IL-2R-module for ALCL biology and identify IL-2Rα-targeting as a promising treatment strategy for ALCL.

S ystemic anaplastic large cell lymphoma (ALCL) is an aggressive CD30-positive, mature T-cell non-Hodgkin lymphoma (T-NHL), which can be divided into two entities according to the presence or absence of Anaplastic Lymphoma Kinase (ALK)-fusions, i.e., ALK-positive (ALK + ) and ALK-negative (ALK − ) ALCL 1,2 . Furthermore, breast implant-associated ALK − ALCL (BIA-ALCL) and primary cutaneous ALCL (pcALCL) are acknowledged as (provisional) entities 2 , in particular the latter with usually good prognosis 3 . Despite implementation of targeted treatment strategies such as CD30 antibody-drug conjugates (ADCs) into up-front treatment regimens 4 , the prognosis of some particular systemic ALCL subtypes is unsatisfactory 5 , and there is an unmet medical need for effective treatment options in the case of relapsed and refractory disease 6,7 .
Super-enhancers (SEs) are genomic regions bound by large clusters of regulatory elements resulting in high-level gene expression of the respective genes, which are marked by extensive stretches of histone 3 lysine 27 acetylation (H3K27ac) 21,22 and help to define unique tumor dependencies and therapeutically relevant vulnerabilities 23 .
Here, we perform a SE analysis by genome-wide chromatin immunoprecipitation sequencing (ChIP-seq) of H3K27ac in ALCL cells in order to characterize in an unbiased manner the transcriptional network and to identify key players that define ALCL pathogenesis. We find BATF3 and IL2R components are among the top SE hits, prompting us to analyze the link between BATF3 and IL-2R components in ALCL cells, and to explore the biological function of the IL-2/IL-2R system in ALCL in detail.

Results
Characterization of the super-enhancers in ALCL. To identify in an unbiased manner cellular dependencies and vulnerabilities which could ultimately be tools for ALCL therapy, we performed genome-wide H3K27ac ChIP-seq for SE detection in 8 ALCL cell lines (5 ALK + , 3 ALK − ). By applying the ROSE algorithm 22,24 , we defined SEs, characterized by extensive H3K27ac, in all 8 cell lines (Fig. 1a, b and Supplementary Fig. 1a, b). These were compared to published H3K27ac ChIP-seq datasets of CD3 + , CD8 + and CD4 + T-cell subsets from healthy donors as well as Tcell-derived Jurkat cell line ( Supplementary Fig. 1b).
Globally, we defined 538 and 722 SEs in Karpas-299 (systemic ALCL, ALK + ) and Mac-1 (derived from cutaneous T-NHL later progressing to ALCL) cell lines (Fig. 1a, b), respectively. 225 genes associated with SEs were shared between both cell lines ( Fig. 1c and Supplementary Table 1). The identified SEs were associated with transcription factors and genes considered key pathogenic factors in ALCL, including CD30, IRF4, JUNB, STAT1 and STAT3 10,25,26 . Furthermore, in the analyzed ALCL cell lines, BATF3 12,27 was consistently found in a SE region with the highest H3K27ac, as were the IL2RA and IL2RB loci (Fig. 1a, b and Supplementary Fig. 1a, b). Remarkably, localization of BATF3 and IL2RA and/or IL2RB in SE regions was a common feature in ALCL cell lines but not in primary T-cell controls or the Jurkat cell line (Fig. 1a, b, c and Supplementary Fig. 1a, b), indicating an ALCL-specific property.
To test whether the SE pattern seen in ALCL cell lines is reflected in ALCL primary patient samples, we performed H3K27ac analysis using the cut-and-run ChIP-seq technique on 2 primary patient samples ( Fig. 1d and Supplementary Fig. 1c). Overall, we found a similar SE pattern as in the ALCL cell lines. In both patients #54 and #208, BATF3 and IL2RB were located in SE regions, and IL2RA was located in SE region in patient #54. Together, these data mirrored our findings in ALCL cell lines, showing the presence of IL2R components and BATF3 in SE regions in ALCL.
IL-2Rα/β/γ are highly expressed in ALCL but not other T-NHL. To comprehensively examine expression of the IL-2R subunits, IL-2Rα (also called CD25), IL-2Rβ (CD122) and IL-2Rγ (CD132), in mature T-NHL, we performed immunostaining of TMA comprising angioimmunoblastic T-cell lymphoma (AITL, n = 8), peripheral T-cell lymphoma not otherwise specified (PTCL-NOS, n = 23), ALCL with (n = 22) or without (n = 23) ALK-translocation, and of non-neoplastic reactive lymph node controls (n = 11) as well as of pcALCL specimens (n = 24). Strong positive staining of all three IL-2R chains was seen in the majority of cases of all ALCL subtypes, with the strongest expression in ALCL, ALK + , whereas no comparable IL-2R expression was observed in the other samples (Fig. 1h, Supplementary Fig. 3a and Supplementary Tables 3 and 4). The BATF3 expression analysis showed a similar pattern (Supplementary Fig. 3b and Supplementary  Table 3), again suggesting a link between BATF3 and IL-2R expression in ALCL. We subsequently tested IL2R and BATF3 expression in a broad cell line panel consisting of ALCL, ALK + , ALCL, ALK − , BIA-ALCL 31 and T-cell control cell lines. Again, expression of IL2RA, IL2RB and BATF3 was restricted to ALCL cell lines ( Fig. 2a and Supplementary Fig. 3c, d, e).
BATF3 directly regulates IL-2Rα/β and CD30 in ALCL. To assess if a direct link exists between BATF3 and IL-2R expression, we examined the mRNA levels of IL2R in BATF3-knockout ALCL cell lines. BATF3 deletion led to significantly reduced mRNA and protein expression of all three IL-2R subunits in all tested ALCL cell lines, and of IL-2Rα and IL-2Rγ in BIA-ALCL cells ( Fig. 2b and Supplementary Fig. 3f, g). Moreover, BATF3 deletion led to reduced growth in ALCL cell lines ( Supplementary Fig. 3h). Protein and pY-protein levels of STAT1, STAT5 and ERK1/2 were not consistently affected by CRISPR/Cas9-mediated BATF3 deletion ( Supplementary Fig. 3i).
Both CD30 and BATF3 display specificity for ALCL 11 , and JUNB is present on the AP-1 site within the CD30 promoter 32,33 . Given our identification of SEs at CD30, BATF3 and IL2R loci, and that CD30 and IL-2R are markers for activated lymphocytes, we tested whether BATF3 also regulates CD30. CD30 expression decreased following BATF3 deletion with a stronger effect in ALCL, ALK + cells (Fig. 2c), and BATF3 was recruited to CD30 regulatory regions as revealed by ChIP (Fig. 2d) and genomewide ChIP-seq (Fig. 2e). To independently assess AP-1 recruitment to regulatory regions of IL2R members in ALCL, we identified AP-1/TRE sites in the promoter or enhancer regions of IL2R loci. The non-conserved 5'-TGAGTAA-3' motif in the IL2RA promoter and the conserved 5'-TGACTCA-3' AP-1/TRE site in the IL2RB enhancer showed strong ALCL-specific DNAbinding ( Fig. 2f and Supplementary Fig. 3j). Recruitment of BATF3 to IL2RA and IL2RB regulatory regions was also observed by ChIP in ALCL cell lines (Fig. 2g), confirming our BATF3 ChIP-seq in Karpas-299 and Mac-1 ALCL cells ( Supplementary  Fig. 2c). Together, our data point towards a direct regulation of IL2R and CD30 genes by BATF3/AP-1 in ALCL.
IL-2 could not be detected in tumor cells of 33 ALCL patients nor in ALCL cell lines ( Supplementary Fig. 4a, b and Supplementary Table 3), suggesting other sources of IL-2 in ALCL. Macrophages and plasma cells that do express IL-2 have been described as being in proximity to ALCL cells 34 , which we did also observe in a number of cases ( Supplementary Fig. 4a). Furthermore, elevated IL-2-levels were previously found in the serum of a subgroup of ALCL patients but none was detected upon remission or in B-NHL controls 35 . These data suggest that IL-2 might be present at least in a subset of ALCL patients and to act on the lymphoma cells.

Targeting IL-2Rα induces ALCL death in vitro and in vivo.
Given the biological role of IL-2/IL-2R in ALCL, the recent development of IL-2Rα-targeting strategies 39,40 , and the medical Fig. 1 Genes encoding BATF3 and subunits of the IL-2R are highly acetylated at H3K27 and positively correlate. a Enhancers were ranked based on increasing H3K27ac signals. Genes within SEs in Karpas-299 (ALK + ) and Mac-1 (ALK − ) ALCL cell lines are marked red. b H3K27ac ChIP-seq tracks at the BATF3 locus with indicated SE regions. c Overlap of genes associated with SEs in Karpas-299 and Mac-1 cell lines. d Enhancers were ranked based on increasing H3K27ac signals. Genes within SEs in primary ALCL patient samples (#54 and #208) are marked in red. e Identification of upstream regulators among deregulated genes in CRISPR/Cas9-mediated BATF3-KO Karpas-299 cells by IPA 2020 analysis. f Analysis of BATF3-correlated genes in a previously published RNA-seq dataset of 23 ALCL patients (BioProject PRJNA255877, SRA identifier SRP044708). Pearson correlation; P value based on t-distribution. g Representative images of 63 patient samples measured of BATF3 and IL-2Rα IHC staining of ALCL, ALK + in paired FFPE tissues sections. h IHC quantification of IL-2Rα, IL-2Rβ and IL-2Rγ expression in mature T-NHL TMAs (reactive lymph node controls, n = 11; AITL, n = 8; PTCL-NOS, n = 23; ALCL, ALK + , n = 22; ALCL, ALK − , n = 23) and pcALCL (n = 24) specimens. P values were determined by two-tailed unpaired Student's t test. All boxwhisker plots represent the median (central line), 25th-75th percentile (bounds of the box) and minimum-maximum (whiskers).
need for new treatment approaches for subgroups of ALCL, we explored IL-2Rα as therapeutic target in ALCL. To this end, we used ADCT-301 (camidanlumab tesirine), a recently developed ADC composed of an IL-2Rα-targeting monoclonal antibody coupled to a highly-potent pyrrolobenzodiazepine (PDB)-dimer cytotoxic warhead 39 . Cells were treated with ADCT-301 or the isotype control ADC B12-SG3249, recognizing HIV-gp120, conjugated to the same linker-warhead as in ADCT-301. At doses of 156 ng/ml and lower, ADCT-301 induced cell death of all IL-2Rα-positive ALCL cell lines while it did not affect viability of IL-2Rα-negative Jurkat and KE-37 cells (Fig. 4c). B12-SG3249 did not affect cell viability. The LD 50 of ADCT-301-induced cell death was remarkably low (between 0.8 and 2.3 ng/ml) for ALCL cell lines, but >156 ng/ml in the IL-2Rα-negative cell lines Fig. 2 BATF3 regulates expression of IL2R subunits and CD30 in ALCL. a IL2RA, IL2RB and IL2RG mRNA expression in cell line panel consisting of 13 ALCL cell lines (6 ALCL, ALK + , 4 ALCL, ALK − and 3 BIA-ALCL), 4 T-cell leukemia-derived control cell lines (T) and PBMCs, or in b CRISPR/Cas9-mediated BATF3-KO and CRISPR control ALCL cells, using quantitative RT-PCR. Data are indicated as means ± SD of biological triplicates. P values were determined by two-tailed unpaired Student's t test. c Cell surface expression of CD30 in BATF3-KO (red) or CRISPR control (green) ALCL cells, isotype control (gray) and unstained cells (black). d BATF3 ChIP of the CD30 regulatory region as compared to a control region (Ref_Ch4). Combined data of two biological replicates are shown as mean ± SEM (see Supplementary Table 8 for ChIP primer sequences). e ChIP-seq tracks at the CD30 locus for BATF3 and H3K27ac in ALCL cell lines with indicated SE regions. f Electrophoretic mobility shift assay (EMSA) of AP-1 TRE complexes on the IL2RA promoter or IL2RB enhancer region (see Supplementary Table 8 for probe sequences) in 10 ALCL cell lines and 4 T-cell leukemia-derived control cell lines. The positions of AP-1 complexes, nonspecific bands (n.s.) and free probes are indicated. Blots shown are representative of three independent experiments with similar results. g BATF3 ChIP on IL2RA or IL2RB regulatory regions as compared to a control region (Ref_Ch4). Combined data of two biological replicates are shown as mean ± SEM (see Supplementary Table 8 for ChIP primer sequences).
( Fig. 4c), indicating the high efficacy and specificity of ADCT-301 for ALCL. Annexin V/propidium-iodide (PI) staining and PARP1-cleavage ( Supplementary Fig. 6b, c) in ADCT-301-treated cells indicated induction of apoptotic cell death. When we tested the effect of ADCT-301 or control ADC B12-SG3249 on STAT1 and STAT5 levels in ALCL cell lines, we found reduction of pYSTAT5 in all tested cell lines and reduction of pYSTAT1 in the FE-PD cell line (Supplementary Fig. 6d).
For determination of ADCT-301 efficacy in vivo, we subcutaneously xenotransplanted three ALCL, ALK − cell lines, Mac-2A, TLBR-1 and FE-PD, into NOD-scid Il2rg null (NSG) mice and intravenously administered a single dose (0.5 mg/kg) of ADCT-301, control ADC B12-SG3249, or vehicle (PBS) when tumors reached 100-150 mm 3 . This single dose of ADCT-301 completely prevented growth of Mac-2A and TLBR-1 tumors and led to tumor growth retardation of rapidly growing FE-PD cells ( Fig. 4d and Supplementary Fig. 6e). The control ADC did not elicit any anti-tumor activity. This remarkable tumor control of ADCT-301 confirmed the high efficacy of IL-2Rα-mediated targeting of ALCL cell growth.

Discussion
ALCL is a heterogeneous disease comprised of systemic ALCL, ALK + and ALK − , pcALCL and BIA-ALCL. Whereas the prognosis of adult ALCL, ALK + and subgroups of ALCL, ALK − is usually good, relapses occur frequently in pediatric ALCL, ALK + , and the outcomes of patients with relapsed/refractory ALCL patients and advanced BIA-ALCL are rather poor, highlighting an unmet medical need for these patients 2,7 .
Ideally, therapeutic targets are not only specific to the respective tumor cells, but, at the same time, are essential components of growth and survival pathways. Screening for SEs is a valuable tool for identification of such genes 21,22,24 . In our SE analyses, the identification of known ALCL-promoting genes, such as JUNB, IRF4, STAT1, STAT3 and CD30 8,9,25,26,41 , validated our approach and independently confirmed previous findings on the pathogenic role of JUNB, IRF4 and STAT transcription factors in ALCL. Of note, BATF3, IRF4 and STAT3 are present in the topten "preferentially essential genes" in ALCL, ALK + cell lines identified by CRISPR screen analyses within the Achilles project (depmap.org; Supplementary Fig. 6f) 42 . Furthermore, BATFfamily and IRF4 genes are involved in chromosomal translocations in ALCL, supporting their biological relevance 29,43 .
Remarkably, both the AP-1-family member BATF3 and IL-2R components were consistently found among the top SEassociated genes in ALCL cell lines and primary ALCL samples, which attracted our attention for detailed analyses given our previous findings on AP-1/BATF function in ALCL and the important role of IL-2/IL-2R for T lymphocyte growth and function 12,18 . We further extend the ALCL pathogenic concept on IL-2/IL-15/IL-2R-components and define the BATF3/IL-2R module as a central component for ALCL growth and survival (Fig. 4e). Furthermore, despite harboring JAK2 or STAT3 mutations 44 , ALCL, ALK − cell lines strongly responded to IL-2, underscoring its growth-supportive role in ALCL. IL-2Rγ was described to be suppressed by NPM1-ALK 20 . However, according to our data (Fig. 1h), IL-2Rγ-silencing rarely occurs in ALCL. Regarding IL-2, its lack of expression in ALCL cells itself may be explained by their strong IL-10 expression, which can actively suppress IL-2 production in T cells 45 . However, cells in the ALCL tumor microenvironment can be assumed to produce IL-2 34,35 , and in the majority of cases, ALCL cells produce IL-15, both accounting for IL-2R stimulation in primary ALCL.
Our data suggest that high-level IL-2Rα expression might be associated with a more aggressive clinical presentation in two independent ALCL cohorts, which has to be confirmed in larger ALCL cohorts. However, this finding reflects our functional studies and is in line with the correlation of high soluble IL-2Rα levels with more aggressive disease and lower EFS at 3 years in ALCL, ALK + children cohorts 35 . As risk factor, IL-2Rα expression might be outcompeted by more intense treatment regimens with brentuximab-vedotin (BV)-containing regimens or SCT in our adult ALCL, ALK − cohort. Moreover, it is unlikely that DUSP22 and TP63 rearrangements, having been described in ALCL, ALK −43,46 , are confounding factors in our survival analysis based on the following findings. In the first clinical report, the 5-year OS of 21 patients with DUSP22 rearrangement was 90% 7 . However, in a later work from the British Columbia Cancer Lymphoid Cancer database, the 5-year OS of 12 patients with DUSP22 rearrangement was 40% 47 . Thus, the prognostic impact of DUSP22 rearrangement is currently unclear, and it may be more related to prognostic factors such as IPI. In addition, in pcALCL, no impact of DUSP22 rearrangement (present in about 30% of cases) on prognosis was observed 48 . Despite that TP63 rearrangement has a confirmed poor prognosis in ALCL, ALK − , it is indeed very rare. We also revealed a BATF3-dependency of Fig. 3 IL-2 and IL-15 promote ALCL proliferation and activate STAT1, STAT5 and ERK1/2. a Indicated 7 ALCL cell lines were treated with rhIL-2 (100 ng/ ml). Control cell lines were SUP-M2 (ALK + , IL-2Rγ-negative) and Jurkat (T-cell leukemia-derived, IL-2Rα-and IL-2Rβ-negative). Resazurine assay was used to measure metabolic activity as a marker of cell proliferation. Data are means ± SD of biological triplicates given as a percentage of the untreated control. P values were determined by two-tailed unpaired Student's t test; n.s. not significant. Structure of IL-2 bound to the trimeric IL-2R complex is shown (PDB ID: 2B5I). b In BATF3-KO cells, no IL-2-induced proliferation in Karpas-299 (ALK + ) and Mac-1 (ALK − ) could be seen compared to CRISPR control. Proliferation was determined by resazurine assay. Data are normalized to untreated control and shown as means ± SD of biological triplicates. P values were determined by two-tailed unpaired Student's t test. c Immunoblot analysis of total STAT1, STAT5 and ERK1/2 proteins and their phosphorylated (pY, pT/Y) forms in Mac-2A (ALK − ), TLBR-2 (BIA) and Jurkat control cells stimulated with rhIL-2 (50 or 100 ng/ml), Neo-2/15 (100 ng/ml) or PBS (−) as control. GAPDH serves as loading control. Blots shown are representative of two independent experiments with similar results in different cell lines. d Limiting dilution of Karpas-299 (ALK + ), Mac-1 (ALK − ) and Jurkat control cells with or without rhIL-2 (100 ng/ml) in a 96-well plate containing RPMI1640 and 10% FCS. At day 0, the indicated number of cells were seeded and incubated for 2 weeks. Representative example of biological triplicates is shown. e Inhibition of IL-2induced proliferation by the IL-2Rα-blocking antibody Basiliximab (1 μg/ml). Proliferation was determined by resazurine assay. Data are normalized to control-treated cells and shown as means ± SD of biological triplicates. P values were determined by two-tailed unpaired Student's t test. f Representative images of 9 patient samples measured of IL-15 IHC staining in fresh-frozen primary ALCL tissue. g Effect of IL-15 (  new effective treatment strategies 49 . With implementation of chemotherapy regimens containing BV, which proved to be the first effective targeted treatment for ALCL in many years 4 , in frontline ALCL therapy, effective treatment options remain lacking in the second line. For ALCL, ALK + patients, pharmacological ALK-inhibition might be a treatment option in this setting 50,51 . IL-2Rα has been investigated as immunotherapeutic tool for other malignancies 52,53 . ADCT-301, which has entered clinical trials (e.g., ClinicalTrials.gov Identifier: NCT02432235 and NCT04052997), killed ALCL cells with high efficacy 39 . Notably, ADCT-301-coupled PBD dimers, given their unique mode of action, are likely to overcome resistance to conventional chemotherapy 54 . Interestingly, ADCT-301 treatment in preclinical models and in patients with advanced solid tumors or lymphomas resulted in transient reduction of T reg cells, while CD4 + and CD8 + T cells were unaffected, which might enhance ADCT-301-mediated anti-tumor activity [55][56][57] . Together, by our unbiased approach we identify the BATF3/IL-2R-module as vulnerability in ALCL, and our data strongly argue for exploration of IL-2Rα-targeting of ALCL in clinical trials. Cell numbers were determined using a CASY Cell Counter. For cytokine stimulation, monoclonal antibody and ADC treatment, rhIL-2 (50 or 100 ng/ml; 1081-IL, R&D Systems), rhIL-15 (50 or 100 ng/ml; 247-ILB, R&D Systems), Neo-2/ 15 36 (100 nglml), basiliximab (1 μg/ml; ab00188, Absolute Antibody) and ADCT-301 (in vitro: 156 ng/ml, in vivo: 0.5 mg/kg; camidanlumab tesirine, ADC Therapeutics) were applied as indicated.

Methods
Human samples. All human samples were obtained with informed written consent and in accordance with the Declaration of Helsinki reviewed by the local ethic boards (Medical University of Vienna 1221/2019; University Hospital Brno 4-306/ 13/1). FFPE tissues from systemic ALCL, ALK − cases collected through TENOMIC, a transnational research consortium on T-cell lymphomas involving several centers in France, Belgium and Switzerland, were used for immunohistochemical validation. The TENOMIC database is approved by the ethical committee "CPP Ile-de-France IX 08-009".
Genome-wide occupancy analysis. ChIP-seq was performed as described 58,59 . Primary antibodies used for analysis are listed in Supplementary Table 7. In short, for each ChIP, 5 μg of antibody was conjugated to 2 mg of M-270 Epoxy Dynabeads (Thermo Fisher) and added to sonicated nuclear extract. Library construction was performed using the Swift 2S Turbo DNA library kit as per the manufacturer's instructions (Swift Biosciences). Illumina sequencing was performed using the NextSeq500 platform with single-end 75 bp reads.
ChIP-seq processing and display. Reads were aligned to the human genome (hg19) using bowtie with parameters -k 2 -m 2 -best and -l set to the read length. For visualization, WIG files were created from aligned ChIP-seq read positions using MACS with parameters -w -S -space=50 -nomodel -shiftsize=200 to artificially extend reads to be 200 bp and to calculate their density in 50 bp bins. Read counts in 50 bp bins were then normalized to the millions of mapped reads, giving reads per million (rpm) values. WIG and TDF files were visualized in the IGV 2.4.6 genome browser.
ChIP-seq enriched regions. Regions enriched in ChIP-seq signal were identified using MACS with corresponding control and parameters -keep-dup=auto and -p 1e−9. The distribution of peaks into other genomic features was determined by overlapping peaks with SEs, typical enhancers, promoters, and gene bodies, in this order. SEs and typical enhancers from the same cell type as the TF ChIP-seq were used. Promoters were defined as 4 kb windows centered on the transcription start sites of RefSeq genes. Gene bodies were defined as the remaining portion of genes longer than 2 kb from 2 kb downstream of the promoter to the transcription end site.
SE identification and assignment. SEs were identified using ROSE (https:// bitbucket.org/young_computation/rose) algorithm with some modifications. Reads overlapping ENCODE blacklist regions were removed using bedtools intersect. From the remaining reads, two sets of peaks of H3K27ac were identified using MACS with parameter sets -keep-dup=auto -p 1e−9 and -keep-dup=all -p 1e−9. H3K27ac peaks were stitched computationally if they were within 12500 bp of each other, though peaks fully contained within + /− 2000bp from a RefSeq promoter were excluded from stitching. These stitched enhancers were ranked by their H3K27ac signal (length * density) with input signal subtracted. SEs were defined geometrically as those enhancers above the point at which the line y = x is tangent to the curve. Stitched enhancers (typical enhancers and SEs) were assigned to the single active gene whose transcription start site is nearest the center of the stitched enhancer. Active genes were determined by taking the top two-thirds of all RefSeq promoters (+/−500 bp) ranked by their H3K27ac signal. H3K27ac signal in promoters was determined using bamToGFF (https://github.com/BradnerLab/pipeline) with parameters -e 200 -m 1 -r -d.
RNA-seq and bioinformatic analysis. RNA isolation was performed using the RNeasy Mini Kit (Qiagen) and RNA quality was assessed on a Fragment Analyzer (Advanced Analytical Technologies) − SS Total RNA 15nt. RNA-seq libraries were prepared using the NEBNext Ultra II Directional RNA Library Prep Kit for Illumina (New England Biolabs) with Poly(A) selection according to the manufacturer's instructions. Library quantification was examined on a Fragment Analyzer − HS NGS Fragment 1-6000 bp and Qubit HS dsDNA Kit (Invitrogen). Libraries were pooled and sequenced to 75 bp single-end on the Illumina NextSeq 500 platform. Sequencing data were analyzed as described 60 .
Immunostaining and quantification. IHC was performed on FFPE (formalinfixed paraffin embedded) sections with the conventional avidin-biotin-peroxidase method. Heat antigen retrieval was performed using citrate buffer pH 6.1 (Agilent Dako). Endogenous peroxidases were quenched by incubating sections in 3% H 2 O 2 in PBS. Sections were blocked using Avidin/Biotin Blocking Kit (Vector Laboratories). Primary antibodies were added in 1% BSA/PBS at 4°C overnight. Slides were incubated with biotin-conjugated secondary antibodies using IDtect Super Stain System-HRP (Empire Genomics) and developed using AEC Substrate Kit (BDPharmingen). Slides were mounted with Aquatex (Merck). Primary antibodies used are listed in Supplementary Table 7  , BATF3 (100 μg/ ml), IL-15 (10 μg/ml) and IL-15Rα (7 μg/ml). Stained slides were further assessed by two experienced pathologists who were blinded to the clinicopathological parameters and patient outcome. Quantification of the slides was determined using the histoscore system. Stained slides were scored for both the intensity and percentage of positively staining cells: 0 negative, 1+ weak, 2+ moderate and 3+ strong staining. Positive staining was considered if present in >1% of cells. The Histoscore was then calculated by multiplying the intensity by the percentage of positively stained cells.
End-point and quantitative RT-PCR. RNA preparation, cDNA synthesis, semiquantitative and quantitative PCR analyses were performed as described 9 . Primers are listed in Supplementary Table 8. EMSA. Nuclear protein preparation and EMSA were performed as described 9 . AP-1 motifs of the respective IL2R promoter (IL2RA) or enhancer (IL2RB) gene regions were identified using LASAGNA-Search 2.0 online tool and EMSA probes NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-25379-9 ARTICLE were designed accordingly. EMSA oligonucleotides are listed in Supplementary  Table 8. Based on the experimental method, no molecular weight markers can be shown for EMSA analyses.
ChIP and real-time PCR analyses. ChIP assays were performed in two biological replicates using antibodies for BATF3 (AF7437, R&D Systems) according to a modified Millipore protocol (http://www.merckmillipore.com/DE/de/product/ ChromatinImmunoprecipitation%28ChIP%29-Array-Kit,MM_NF-17-295#anchor_BRO) as described 12 . In short, cells were fixed using the two-step cross-linking method (2 mM disuccinimidyl glutarate for 30 min followed by 1% formaldehyde for 5 min) and lyzed with 50 mM Tris-HCl, pH 8.0/5 mM EDTA/1% SDS. Thereafter, samples were sonicated with the Bioruptor Plus (Diagenode; 13 cycles, intensity High, sonication 30 s/break 30 s per cycle). Chromatin was precleared with bovine serum albumin (BSA)-saturated Protein A-Sepharose and incubated overnight at 4°C with BATF3 antibody (27 μg/1 × 107 cells). Immunocomplexes were collected with BSA-saturated Protein A-Sepharose for 1 h at 4°C. After washing, protein-DNA complexes were eluted using 1% SDS/0.1 M NaHCO 3 . Reversal of the cross-linking, RNAse treatment, proteinase K digestion, and DNA purification by phenol-chloroform extraction were performed according to standard protocols. Primary antibody and primer sequences are shown in Supplementary Table 7 and Table 8, respectively. Quantitative PCR was carried out in technical triplicates with ChIP-DNA isolated from to 3 × 10 5 cell equivalents using the CFX96 system and GoTaq qPCR Master Mix (Promega). Non-recruiting intergenic regions on chromosomes 4 (Ref_Ch4) were amplified as references. A total of 4 ng input DNA was used as a control.
Cell proliferation assay. Cells were starved in medium with 1% FCS overnight and seeded in 96-well plates with 2 × 10 4 cells per well. Cell proliferation was measured after 72 h of cytokine stimulation with or without monoclonal antibody treatment by resazurin assay or [ 3 H]-thymidine incorporation according to manufacturer's instructions. All values were normalized to the control-treated cells.
Immunoblotting. Whole protein lysate preparation and immunoblot analyses were performed as described 28 . Primary antibodies used are listed in Supplementary  Table 7  Murine xenograft model. Both male and female NSG mice were purchased from the National Cancer Institute (Frederick, MD). ALCL cells (1 × 10 6 ; FE-PD, Mac-2A or TLBR-1) were subcutaneously implanted into the hind flanks of nine-to ten-week old mice. When tumor volumes reached~100-150 mm 3 , mice were randomly allocated into three groups to receive vehicle (PBS), non-binding ADC (B12-SG3249) or ADCT-301 (camidanlumab tesirine) (both ADC Therapeutics, 0.5 mg/kg, intravenous injection). Tumors were measured with caliper every three days. Tumor size was estimated using the following formula: volume (mm 3 ) = (length × width 2 )/2. Each mouse was euthanized when its tumor reached the end-point volume of 2000 mm 3 or at the study end. Animal work was carried out according to an ethical animal license protocol that was approved by the Medical University of Vienna and Austrian Ministry of Education and Science (BMBWF-66.009/0200-V/3b/2018 and Addendum Zl. 18/115-97/98, 2019).
Statistical analysis. Statistical analyses were performed using GraphPad Prism v6.0 and R v4.0.3 software. Data of protein expression in TMAs are presented in box-and-whisker plots with indication of medians, quartiles and ranges, each data point represents an individual score. Correlation of data was determined by the Pearson correlation coefficient. Other data are presented as means ± SD or SEM as stated in the respective figure legends. Differences between mean values of two groups were evaluated by the two-tailed unpaired Student's t test. Estimation of EFS and OS was performed using the Kaplan-Meier method using IBM SPSS Statistics Version 22 with the survival package of R (v3.2-7; Therneau, 2020). Events were defined as relapse, progressive disease or death from any cause. Differences between groups were compared by a log-rank test. P values < 0.05 were considered to be statistically significant.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
All RNA-seq and ChIP-seq datasets produced in this study are deposited in Gene Expression Omnibus (GEO) under SuperSeries GSE158916. Source data is provided for the Fig. 1a-d, Supplementary Fig. 1 and Supplementary Fig. 2a-d. The Crescenzo et al. 29 and Iqbal et al. 30 publicly available data used in this study are available in the Sequence Read Archive (SRA) and GEO database under accession code SRP044708 and GSE19069, respectively. ChIP-seq of H3K27ac and input DNA of normal T-cell subsets and Jurkat cell line used in this study are publicly available in GEO under the accession numbers GSM1058764, GSM1058789, GSM772835, GSM772916, GSM1102781, GSM1102806, GSM1296384 and GSM569086. The remaining data are available within the Article, Supplementary Information or Source Data file. Source data are provided with this paper.

Code availability
Computer code used in this study is available within Supplementary Software.