NPM–ALK-Induced Reprogramming of Mature TCR-Stimulated T Cells Results in Dedifferentiation and Malignant Transformation

Translocation between the genes encoding anaplastic lymphoma kinase (ALK), a receptor tyrosine kinase physiologically expressed in immature neuronal tissue, and nucleophosmin (NPM), a ubiquitously expressed nucleolar protein, generates the NPM–ALK fusion protein, which is a key driver of anaplastic large-cell lymphoma (ALCL; refs. 1, 2). NPM–ALK forms oligomers, resulting in constitutive activation of its kinase domain, which is required for NPM–ALK-mediated transformation of T lymphocytes (3). Extensive mutagenesis studies of tyrosine-based motifs within NPM–ALK's protein sequence followed by immunoprecipitation studies have revealed key amino acid residues by which a number of signaling proteins bind to NPM–ALK including IRS1, SRC, SHC1, SHP2, GRB2, SOS, and PLCγ (4). For example, using rodent fibroblasts as a model, the Y567 and Y664 sites were found to be dispensable for cell transformation (3). However, another study using murine leukemic Ba/F3 cells demonstrated that mutation of the Y664 residue in NPM–ALK failed to result in NPM–ALK-mediated transformation due to loss of PLCγ recruitment (5). To date, studies on NPM–ALK signaling have been performed using patient tumor samples and cell lines or non-NPM–ALK-expressing cell lines that have undergone transformation with NPM–ALK expressed exogenously (4, 6). Thus, it is not clear which, if any, of these pathways are principle drivers that lead a primary human T cell to become an ALCL.

The NPM–ALK kinase is able to activate several downstream signaling cascades, including STAT3, cJUN/JunB/AP1, PI3K/AKT/mTOR, MEK–ERK, and NFAT (6, 7), but which of these signaling pathways are induced early and which contribute to transformation are unknown. Intriguingly, the NPM–ALK translocation can also be detected in T cells of some healthy individuals and cord blood (8–11), suggesting that expression of an oncogenic tyrosine kinase (NPM–ALK) is not sufficient to induce transformation in these cells, but rather that additional, yet undefined, signals are necessary (12). Because ALK is not normally expressed in mature T cells, a cell type that is generally resistant to transformation (13), it is remarkable that this aberrantly expressed kinase is able to reprogram T cells to become tumors. Understanding which pathways are altered early in the transformation process will provide key clues to delineate how NPM–ALK expression functions to transform T cells.

Historically, mature T cells were thought to be the cell of origin for ALCLs because the transformed cells have rearranged T-cell receptor (TCR) loci and produce cytotoxic molecules like perforin and granzyme B (14–16). However, other studies suggest that a less differentiated cell may be the cell of origin for ALCL based on cell morphology and methylation patterns (17–19). Transgenic mouse models of NPM–ALK transformation have attempted to address some of these unresolved issues, but these have failed to produce tumors that recapitulate primary human ALCL (20). Thus, it has been difficult to study NPM–ALK-driven oncogenesis in T cells. We previously demonstrated that introducing NPM–ALK into activated human CD4 T cells mediated their transformation into ALCL lines indistinguishable from those isolated from patients (21). Here, we use this primary human T-cell model of NPM–ALK-mediated transformation to study mechanistically how NPM–ALK drives T-cell tumorigenesis. We show that NPM–ALK rapidly alters gene expression in mature T cells, driving them to a less differentiated state that genetically resembles a more thymic-progenitor–like state. These findings support the notion that a mature T cell can be the cell of origin for ALCL. We identify critical tyrosine signaling residues and signaling pathways whose disruption blocks T-cell transformation. We further demonstrate that NPM–ALK expression alone is not sufficient to transform primary human CD4 T cells, but rather that TCR engagement is also required. Our data provide a time line of the changes a T cell undergoes to become an ALCL, and a new understanding of the requirements emanating from both NPM–ALK and the T-cell intrinsic signaling machinery that lead to transformation.

The NPM–ALK constructs pTRPE–NPM–ALK and pTRPE–NPM–ALK^KD (K210R) were used to produce lentivirus for NPM–ALK and NPM–ALK^KD expression and have been previously described (21). To create single mutants of NPM–ALK, the QuikChange II Site-Directed Mutagenesis Kit (Agilent Technologies) was used to perform site-directed mutagenesis of the indicated single mutants of NPM–ALK. For some NPM–ALK single mutants and NPM–ALK mutants containing two or three mutations, G-block gene fragments (IDT) were used and standard molecular biology was used to introduce these fragments into pTRPE. Sequences encoding the high affinity IV9-specific TCR were provided by Bent Jakobsen (Adaptimmune Inc) based on a previously described natural TCR (22). Supernatants containing lentiviral vector particles were produced as described previously (23).

Generation of ALCL lines by NPM–ALK transduction was previously described (21). In brief, deidentified primary human CD4 T cells were obtained from the Human Immunology Core at the University of Pennsylvania (Philadelphia, PA) under an Institutional Review Board–approved protocol. These T cells were activated with CD3/28-coated beads (Thermo Fisher Scientific), transduced with NPM–ALK lentiviral vectors, counted, and reduced to 0.5 million per mL every two days. In some experiments, freshly isolated T cells were incubated in 100 U IL2 (Proleukin/Aldesleukin from Clingen), 10 ng/mL IL7 (R&D Systems), and 5 ng/mL IL15 (BioLegend) for 4 days prior to being transduced with NPM–ALK-expressing vectors once a day for the next 3 days. T cells and ALCL lines were maintained in Gibco complete RPMI1640 (Thermo Fisher Scientific) supplemented with 10% FBS (Seradigm), 2 mmol/L Glutamax (Thermo Fisher Scientific), penicillin 100 U/mL/streptomycin 100 μg/mL (Thermo Fisher Scientific), and 10 mmol/L HEPES (Thermo Fisher Scientific). Karpas 299, DEL, SUP-M2, and SUDHL1 ALK⁺ ALCL cells were a kind gift from Megan Lim and Kojo Elenitoba-Johnson (University of Pennsylvania, Philadelphia, PA). These human ALK⁺ ALCL cell lines were maintained in complete RPMI1640. K562 cells expressing HLA-A2, CD64, and the IV9 minigene that contains 50 amino acids surrounding the IV9 epitope were constructed as described previously (24).

Sites of integration of lentiviral vectors were determined essentially as described previously (25–27). Briefly, DNA was purified from cells, then fragmented by sonication. DNA linkers were then ligated to the broken DNA ends, and PCR was carried out using primers that bind to the linker and the vector LTR. PCR products were sequenced using the Illumina method. Reads were aligned to the hg38 draft of the human genome. Clone size was inferred by counting the numbers of sites of ligation of linkers recovered for each unique integration site (25). The distributions of vector integration sites were compared with distributions of random sites using the ROC area method (28).

To enrich for NPM–ALK-expressing cells, cells were incubated with anti–PD-L1-APC (BioLegend, 329708), followed by incubation with anti-APC MicroBeads (# 130-048-801, Miltenyi Inc) and purification using Miltenyi MS columns as per the manufacturer's instructions. Untransduced and NPM–ALK^KD-transduced T cells were purified similarly using anti–CD4-APC (BioLegend, #300537) instead of anti–PD-L1-APC. RNA was extracted using the Qiagen RNeasy Kit (Qiagen, 74104) according to the manufacturer's instructions. 0.5 to 2.0 million cells were used for each extraction. Cells expressing NPM–ALK, NPM–ALK^KD, NPM–ALK^Y664F, and NPM–ALK^TrpM were separated using Miltenyi separation as described above on days 6, 9, and 12. Samples from three independent donors were used for RNA sequencing (RNA-seq) on days 6 and 9, and from two donors on days 12, 33, and 63. Samples from day 33 and 63 were fully transformed, and so PD-L1/CD4 cell sorting was not required. Because NPM–ALK^KD and untransduced cells do not survive beyond day ∼15 to 20, NPM–ALK^KD cells were used for comparison in NPM–ALK sequencing to NPM–ALK on days 33 and 63. Libraries for RNA-seq were prepared by Beijing Genomics Institute (Hong Kong, China) using 100 ng of purified total RNA and the Truseq stranded total RNA RiboZero Library Preparation Kit (Illumina, catalog no. RS-122–2201) strictly following the Illumina guide (15031048 E). Briefly, rRNA was depleted with rRNA Removal Mix (RRM) and then fragmented into approximately 160-bp fragments. rRNA-depleted RNA fragments served as templates for first-strand cDNA synthesis using random hexamer primers, followed by second-strand synthesis with the addition of buffer, dNTPs, RNase H, and DNA polymerase I. Double stranded cDNA was purified using the QiaQuick PCR extraction kit (Qiagen) followed by end repair, base A addition, and ligation of sequencing adapters. Ligated fragments were purified by magnetic beads and amplified via PCR. The resulting library products were quantified with the Agilent2100 bioanalyzer and were sequenced using an Illumina HiSeqTM 4000 for a total of 4 GB of 50-bp paired-end read data per sample. RNA-seq gene counts were determined with the Salmon software package (29) using the hg38 gencode29 transcriptome release, after which, counts were normalized and quantified with the DEseq2 software package (30). All analysis software developed for this study is available via the Zenodo data server (DOI:10.5281/zenodo.4477729).

Cells were washed with PBS, and stained with Fixable Viability Dye eFluor-780 (eBioscience) as per the manufacturer's instructions. Cells were then washed with FACS wash (2.5% FBS/PBS) one time, and surface antibodies were added (NPM–ALK; Miltenyi clone REA425, catalog no. 130-106-446), CD4 (BD Biosciences, clone RPA-T4 or BioLegend, clone OKT4), CD274 (BD Biosciences, clone MIH1), CD30 (BD Biosciences, clone BerH8). The A2-IV9 tetramer was provided by NIH Tetramer Facility. Cells were washed again and fixed with Invitrogen Fixation Medium (Medium A) for 15 minutes at room temperature or overnight at 4°C. Cells were then washed once with FACS wash, and intracellular staining antibodies were added in Permeabilization Medium (Medium B; Invitrogen/Thermo Fisher Scientific). Intracellular staining was performed for 30 minutes, followed by one or two washes with FACS Wash. Cells were then fixed with 2% PFA. Data acquisition was performed using an LSR II or Fortessa, and data analysis was performed using FlowJo. Intracellular cytokine staining was performed as described previously (31). Where indicated, T cells where incubated with 3 μg/mL PMA (Sigma Aldrich, P1585), 1 μg/mL ionomycin (Sigma Aldrich, 407950), and Brefeldin A (GolgiPlug; BD Biosciences, 555029), and then incubated at 37°C for 4 hours. Cells were stained for surface markers and viability, then fixed using Fixation Medium A (Thermo Fisher Scientific, GAS001S100), washed, and a final intracellular cytokine staining was performed in Permeabilization Medium B (Thermo Fisher Scientific, GAS002S100) using antibodies detecting IL2, TNF, GM-CSF, and IFNγ for 30 minutes at room temperature. Cells were then washed and fixed in 4% PFA and analyzed using a BD LSR II Flow Cytometer. For CTV dilution assay, 10 million primary CD4 cells were washed twice with PBS and resuspended in 5 μmol/L of CTV for 20 minutes at room temperature in the dark. Following washout, T cells were stimulated in 100 U IL2 (Proleukin/Aldesleukin from Clingen), 10 ng/mL IL7 (R&D Systems), and 5 ng/mL IL15 (BioLegend) or CD3/28-coated beads (Thermo Fisher Scientific).

Genomic DNA was prepared using the QIAgen DNeasy Blood and Tissue Kit according to the manufacturer's protocol. The genomic DNA was then modified by bisulfite treatment using the EZ DNA Methylation Kit (Zymo Research) as per the manufacturer's instructions. Bisulfite converted DNA was PCR amplified using primers specific for the promoters of the genes indicated. For the SHP-1 promoter loci, bisulfite sequencing was performed as described previously (32, 33). For LCK, CD3ϵ, and TCF7 bisulfite specific primers that amplify both methylated and unmethylated DNA were used. Bisulfite primer sequences used were as follows: LCK forward (TAGGTGGGAGGTAGGTAGTGTAGTT), LCK reverse (TTCTCTCTAAACCTCAAAAACCTCAT); CD3ϵ forward (TTTTTGTTTTATTTTGGGTTTTTTA), CD3ϵ reverse (ACTCCTTACAAAAATCACATCCTAC); TCF7 forward (GGATAAATTTTTAGAGTTTTTGGAGG), and TCF7 reverse (ACTCTCCACCCCTAAATTCTACC). The PCR product sizes and number of CpG sites spanned for each gene were: LCK: 214 base pairs with 5 CpG sites; CD3ϵ: 206 base pairs with 4 total and 2 readable CpG sites spanned; and TCF7: 168 base pairs and 14 CpG sites spanned. Primers were created using MethPrimer (http://www.urogene.org/methprimer). For DNA sequence analysis, PCR products obtained with the indicated primer pairs were gel purified using the Qiagen Gel Extraction Kit. Purified DNA was sequenced using the indicated reverse primer within each primer pair and sequenced/analyzed on an automated DNA sequencer at the University of Pennsylvania DNA Sequencing Facility. Chromatograms were analyzed using the Vector NTI software (Thermo Fisher Scientific). At the CpG sites indicated, the percentage of Cs was calculated from the chromatogram readout.

Twenty to 100 million of NPM–ALK or NPM–ALK-mutant–expressing T cells were resuspended at a concentration of 25 million cells/mL in 1X RIPA buffer (Cell Signaling Technology) and 1 mmol/L phenylmethylsulfonylfluoride (Cell Signaling Technology). The lysate was then isolated by centrifugation (10 minutes at 14,000 × g) and supernatant extracted. Cell lysate supernatant protein concentration was determined using the Pierce Coomassie Protein Assay Reagent Kit (Thermo Fisher Scientific). A predetermined titrated amount (0.5–5 μg) of the NPM–ALK polyclonal goat antibody (Santa Cruz Biotechnology) was added and Protein G Dynabead/Antibody (Thermo Fisher Scientific) were used as recommended by the manufacturer to isolate complexes that bound to the respective NPM–ALK proteins. The samples were run at 120 V into a 1-mm, 10% polyacrylamide 10-well gel for 1 hour along Precision Plus dual color protein standards using NuPAGE MES SDS running buffer (Bio-Rad) and stained with SimplyBlue Safestain (Thermo Fisher Scientific) according to the recommended protocol. Bands were excised and Tandem LC/MS-MS analysis was performed as described previously (34) at Proteomics & Metabolomics Facility at the Wistar Institute (Philadelphia, PA).

Previously, we demonstrated that primary human T cells transduced with a lentiviral vector expressing NPM–ALK generate tumors indistinguishable from primary, patient-derived ALCL tumors (21). To study the step-by-step changes these T cells undergo as they become ALCLs, we generated 29 unique NPM–ALK-expressing tumor cell lines from primary human CD4 T cells (Fig. 1A). Consistent with our previous findings, full T-cell transformation took place approximately 30 days postintroduction of NPM–ALK, whereas untransduced cells and T cells transduced with a kinase-defective (K210R) NPM–ALK (NPM–ALK^KD) returned to resting cell size (Supplementary Fig. S1A) and started to die after approximately 15 days in culture (21, 35). We observed extensive cell death and a decrease in the number of wild-type NPM–ALK-expressing cells between days 6 and 20 that was not observed in T cells transduced with NPM–ALK^KD (Fig. 1B and C; Supplementary Fig. S1B), suggesting that expression of NPM–ALK induces oncogenic toxicity, paralleling previous studies (36, 37).

To capture the temporal changes in gene expression induced by NPM–ALK as it drives T cells to become malignant, we analyzed the gene expression profile of three of these lines at 6, 9, 12, 33, and 63 days after induction of NPM–ALK expression. Because we compare expression profiles with similarly activated T cells transduced with NPM–ALK^KD, our analysis focuses on the changes induced by kinase active NPM–ALK expression within the background of those induced by CD3/28 costimulation (38). To determine whether our analysis mirrored data from previous studies that examined gene expression in patient-derived ALCL, we first examined transcripts known to be regulated by NPM–ALK (32, 39–42). Consistent with these previous studies, we saw upregulation of GZMB, IL10, and PDL1 (CD274) and downregulation of ZAP70, CD3E, and LCK in the ALCL lines we generated (Fig. 1D). In addition, we took advantage of a 24-gene signature that was developed to distinguish ALK⁺ ALCL from related T-cell tumors with a sensitivity and specificity of greater than 95% (43). This comparison suggests that the T-cell tumors generated in vitro via NPM–ALK transduction have many of the characteristics of patient derived ALCL (Fig. 1E). Principal component analysis of the RNA expression data revealed that the NPM–ALK-expressing T cells cluster together in a time point–dependent manner, while the NPM–ALK^KD and untransduced T cells cluster tightly with each other and separately from the NPM–ALK-expressing cells (Fig. 1F). The distribution of time points for the NPM–ALK samples within the principal component analysis (PCA) plot suggests a stepwise progression of the cells from normal CD4 T cells to transformed cells on days 33 and 63 (no regression is observed). The proximity of the day 33 and 63 samples also suggests that approximately 5 weeks after the introduction of NPM–ALK, most gene expression changes induced by NPM–ALK have been set and do not change significantly afterwards. Comparison of gene expression patterns in NPM–ALK- and NPM–ALK^KD-expressing cells demonstrates that NPM–ALK has dramatic effects on expression of many genes (>5,000) as early as 6 days after it is expressed (Fig. 1G). At these early time points, fold change increases in genes upregulated by NPM–ALK are larger than those in downregulated genes. At later time points, an equal number of genes are up and downregulated, and fold changes in genes both induced and repressed by NPM–ALK become of similar magnitude, suggesting that repression of some genes may reflect secondary regulatory events. Further analysis of mRNA expression within 10-gene clusters revealed coordinated waves of gene expression changes over time (Supplementary Fig. S2A–S2C). Through gene set enrichment analysis (GSEA; Fig. 1H), and using annotation based on Kyoto Encyclopedia of Genes and Genomes (KEGG; Supplementary Fig. S2D), we noted perturbation of many signaling pathways as a healthy T-cell becomes a tumor. After six days of NPM–ALK expression, we observed a striking repression of several classes of genes targeted by c-MYC, E2F, mTOR, and G2M signaling (Fig. 1H and I). However, three to six days later these pathways switched from being repressed to being activated (Fig. 1H, bottom). Other pathways such as the STAT3 signaling and epithelial-to-mesenchymal transition (EMT) programs remained upregulated at all time points, while those associated with DNA repair and WNT/β-catenin signaling were repressed at all time points. These studies highlight the coordinated, dynamic, and profound changes in gene expression profiles that occur during oncogenic CD4 T-cell transformation.

One striking observation from the gene expression data is that many T-cell–specific transcription factors, surface receptors, and other functional molecules were strongly repressed by NPM–ALK, while several genes known to be expressed in immature lymphocyte progenitors, hematopoietic progenitor cells, and non–T-cell lineages were upregulated (Fig. 2A–E). These data suggest that NPM–ALK induces a dedifferentiation of T cells toward a progenitor-like state. A previous study demonstrated that ALK⁺ ALCL patient cells have DNA methylation patterns similar to early thymic progenitors (17). Therefore, we examined the methylation status of T-cell–specific genes to determine if NPM–ALK expression can induce methylation of specific gene loci, which are epigenetically silenced in thymic progenitors. Indeed, we detected increased levels of methylation at the promoters of T-cell–specific proteins in cells expressing NPM–ALK, including CD3E, LCK, TCF7, and SHP1 promoters (Supplementary Fig. S3A–S3C). To determine the kinetics of DNA methylation, we analyzed the methylation status of the SHP-1 and LCK promoters over time as T cells were undergoing NPM–ALK-induced transformation (Fig. 2F and G). Methylation of the SHP-1 promoter was not detected until after the T cells had become fully transformed, with limited methylation of the promoter seen on day 56 but gradually increasing afterwards up to day 241. Similarly, no methylation was observed at the LCK promoter on days 11 and 28 after NPM–ALK expression; however, methylation appeared by day 56 and had increased by day 118. Because the RNA-seq data demonstrate downregulation of LCK and SHP1 RNA as early as day 6 after introduction of NPM–ALK, these data demonstrate that methylation occurs after expression of these proteins is inhibited, consistent with previous observations that DNA methylation is not an early step in gene silencing, but rather is important for reinforcing and maintaining genes in the off state after the initial silencing event (44). These data reveal the potent transforming abilities of NPM–ALK and the remarkable plasticity of mature CD4 T cells.

Our expression profiling data show that NPM–ALK rapidly alters the expression of genes involved in TCR signaling and T-cell identity. This prompted us to investigate how rapidly NPM–ALK alters a T cell's ability to respond to cognate antigen. For this purpose, we cotransduced CD4 T cells with NPM–ALK or NPM–ALK^KD together with a TCR that confers CD8-independent recognition to an HLA-A2–restricted, HIV DNA polymerase–specific antigen termed A2-IV9 (Fig. 3A; ref. 22). Next, we cocultured these T cells with K562 cell–based artificial APCs engineered to express HLA-A2 and a portion of the HIV DNA polymerase corresponding to the A2-IV9 antigen linked to GFP (KTA2.IV9) and measured intracellular cytokine production (Fig. 3B). NPM–ALK^KD, A2-IV9–specific expressing T cells produced high levels of TNFα and IL2. In contrast, NPM–ALK, A2-IV9–specific expressing T cells had significantly impaired ability to produce these cytokines after antigenic stimulation, even after treatment with PMA and ionomycin. We next investigated how rapidly NPM–ALK alters cytokine production by stimulating T cells with PMA and ionomycin 1, 3, 5, and 7 days after transduction with NPM–ALK or NPM–ALK^KD (Fig. 3C and D). One day after transduction of NPM–ALK into the cells (day 2 of culture), before NPM–ALK reaches its maximum levels of expression, we noted some inhibition of cytokine synthesis. Between 3 and 7 days after the introduction of NPM–ALK (4 and 8 days of culture), we observed significantly decreased levels of GM-CSF, TNF, and IL2 with repression of IL2 and GM-CSF being the most striking. We observed elevated levels of IFNγ production in NPM–ALK cells as compared with NPM–ALK^KD cells after treatment with PMA and ionomycin at these time points, consistent with an increased IFNγ signaling signature (Fig. 1H). Together, these data indicate that most but not all of the T-cell effector cytokines produced in response to antigen stimulation are rapidly repressed by NPM–ALK expression.

To better understand the NPM–ALK-generated signals that are required to mediate transformation, we generated a large array of tyrosine mutants based on previous studies that investigated NPM–ALK-transformed cell lines (Fig. 4A; ref. 4). Surprisingly, no single Y to F mutation interfered with the ability of NPM–ALK to transform T cells (Fig. 4B), and several of these mutations (Y461F, Y156F, and Y664F) appeared to promote T-cell transformation. However, only one of these mutants (NPM–ALK^Y664F) consistently transformed T cells faster that NPM–ALK (Fig. 4C; Supplementary Fig. S4A). In NPM–ALK^Y664F-transformed cells we observed less NPM–ALK-mediated cell death and increased TCR diversity when compared with NPM–ALK-transformed cells (Supplementary Fig. S4B–S4D). This suggests that this mutant, when expressed in primary human T cells, causes less initial oncogenic stress, resulting in more individual T cells becoming transformed. Gene expression profiling comparing T cells transduced with NPM–ALK and NPM–ALK^Y664F revealed similar changes in gene expression relative to NPM–ALK^KD T cells as indicated by PCA dissimilarity and GSEA analysis (Fig. 4D and E), but the degree to which the related pathways are regulated appear to be less pronounced when compared with T cells expressing NPM–ALK (compare Fig. 1H vs. Fig. 4E). In addition, a set of genes are discordantly regulated between T cells expressing NPM–ALK and NPM–ALK^Y664F (Supplementary Fig. S4E), suggesting that altered gene expression and signaling of the NPM–ALK^Y664F mutant in comparison with wild-type NPM–ALK may also contribute to its enhanced ability to transform primary human T cells.

Because no single tyrosine mutant blocked the ability of NPM–ALK to transform T cells, we created a series of double and triple Y to F mutants within NPM–ALK to define the fewest number of mutations required to block NPM–ALK's transforming ability. After evaluating a number of combinations (Supplementary Fig. S4F), we identified a triple point mutant, NPM–ALK–Y567–644–646F (NPM–ALK^TrpM) that failed to transform mature CD4 T cells (Fig. 4F; Supplementary Fig. S4G). Importantly, the NPM–ALK^TrpM mutant induced some signaling in T cells as demonstrated by induction of CD30 expression at early time points (Fig. 4G, day 8), but this was not maintained and the cultures were not viable after 20 to 30 days after NPM–ALK^TrpM transduction. Thus, the NPM–ALK^TrpM mutant can temporarily display some of the hallmarks of NPM–ALK-mediated transformation but is unable to fully support and sustain T-cell transformation.

To define differential gene expression patterns induced by NPM–ALK^TrpM relative to NPM–ALK, we performed RNA-seq on cells expressing NPM–ALK, NPM–ALK^KD, or NPM–ALK^TrpM after 6 and 9 days of culture. PCA revealed that cells expressing NPM–ALK^TrpM cluster between NPM–ALK and NPM–ALK^KD (Fig. 5A), which is consistent with our previous observations showing that NPM–ALK^TrpM initiates aspects of NPM–ALK signaling, but cannot mediate full T-cell transformation (Fig. 4F and G). GSEA comparison between NPM–ALK^TrpM and NPM–ALK^KD showed that some gene sets including E2F, IFNα, and IFNγ, and KRAS are similarly regulated by both NPM–ALK and NPM–ALK^TrpM (Fig. 5B). However, there were a number of pathways that are regulated by NPM–ALK but not by NPM–ALK^TrpM. For instance, compared with the changes in regulation happening in NA versus NA-KD (Fig. 1H), we did not see induction of genes regulated by EMT, STAT3, or Notch nor did we observe downregulation of MYC, mTORC-1 and WNT signaling pathways at days 6 and 9. We also compared gene sets identified in Fig. 2 that focus on T- and stem cell–specific factors in NPM–ALK^TrpM- versus NPM–ALK^KD-expressing T cells (Fig. 5C). NPM–ALK^TrpM was able to downregulate T-cell–specific surface receptors similarly to NPM–ALK but not T-cell–specific transcription factors or signaling proteins. We did not see induction of the non–T-cell–specific genes or genes associated with embryonic stem cells by NPM–ALK^TrpM, further supporting the GSEA analysis indicating that NPM–ALK^TrpM is unable to induce the dedifferentiation program that may be required to enable transformation of mature T cells.

To gain further insight into the signaling complexes that are required for NPM–ALK-mediated transformation, we identified proteins whose binding to NPM–ALK is lost in NPM–ALK^TrpM. We performed immunoprecipitation of NPM–ALK followed by Tandem LC/MS-MS from T cells expressing either NPM–ALK, NPM–ALK^KD, NPM–ALK^Y664F, or NPM–ALK^TrpM. We performed immunoprecipitation on day 8, which is a time of high cell viability and also when CD30 and PD-L1 are still expressed in NPM–ALK^TrpM T cells. NPM–ALK^TrpM failed to bind several signaling proteins such as SOS2, GRB2, and SHC1 while others such as STAT1 bound NPM–ALK and NPM–ALK^TrpM equivalently (Fig. 5D; Supplementary Table S1). These proteins have been shown to bind NPM–ALK in patient-derived ALCLs or cell lines forced to express NPM–ALK (45, 46). We also identified several proteins not previously demonstrated to bind to NPM–ALK, including SHC4, ARHGEF5, and UBASH3B, which interact with NPM–ALK but fail to interact with either NPM–ALK^KD or NPM–ALK^TrpM. The SHC1 and UBASH3B interactions with NPM–ALK, but not NPM–ALK^TrpM, were further validated by Western blotting (Supplementary Fig. S5). Thus, by making mutations to three residues within NPM–ALK outside of the kinase domain, we no longer observed NPM–ALK-mediated transformation, indicating that complexes that bind to these motifs and the pathways they regulate provide signals that are critical for T-cell transformation.

Lentiviral insertional mutagenesis has been a potent means to discover genes that potentiate CAR T-cell function (47, 48). We wished to determine whether we could use this approach to identify genes that promoted NPM–ALK transformation. To determine whether T cells harboring specific integrations were enriched over time using the samples described in Fig. 1A, we first analyzed samples from three experiments collected at approximately 12, 28, and > 50 days post NPM–ALK transduction. We observed that the number of independent clones, as reported by marking with unique vector integration sites, decreased over time until only a handful of clones dominated the culture (Fig. 6A and B; Supplementary Fig. S6A). Analysis of the distributions of integration sites relative to chromosomal features showed that integration was favored near active transcription units and associated epigenetic marks (Fig. 6C and D), and that there were no major differences early versus late, or comparing NPM–ALK with NPM–ALK^Y664F cells.

Next, we hypothesized that the repeated isolation of integration sites in or near the same gene may indicate that insertional mutagenesis altered activity of the targeted gene, thereby promoting cell growth. We thus examined all 29 transformation experiments to determine whether any signaling pathways or individual genes were consistently targeted by lentiviral integration in NPM–ALK tumors that expanded. There was great heterogeneity in the number of clones within a particular cell line, with some tumor lines being nearly homogenous with a single vector insertion event, while others contained well over 20 separate vector integrations (Supplementary Fig. S6B). Integration acceptor sites were spread throughout the genome and were not preferentially located in oncogenes (Supplementary Table S2). However, we did observe insertion of lentiviral vectors near a distinct set of functionally diverse genes in multiple independent samples (Fig. 6E and F), including genes involved in chromatin structure (ARID1A, INO80), cytoskeleton (ACTN4, ACTR3, ARAP2, ATXN7), RNA turnover (YTHDF3, ABCE1, DNX9), protein translation (E1F3A, E1F3H), tumor suppression (MOB1A), and ubiquitination (AR1H2, UBE2G1).

We carried out a permutation test to assess the significance of the repeated isolation of integration sites in genes associated with expanded clones. We defined expanded clones as those comprising at least 1% of all cells sampled, yielding a total of 295 clones. We then drew 295 clones from the pool of all clones 1,000 times, and asked how often we obtained multiple integration sites in the same locus. Of the thousand draws, only 15 showed 5 isolations within the same gene, suggesting the detection of 5 expanded clones at ARID1A was likely not due to chance (simulation P value of 0.015). Lower numbers of recurrences were obtained more frequently and could be due to chance. Together, these analyses provide the rationale to perform mechanistic studies to determine the role ARID1A plays in NPM–ALK-mediated T-cell transformation.

Healthy individuals can harbor resting T cells that have the NPM–ALK translocation (8–11). We wished to see whether we could recapitulate this in our system to interrogate why some individuals progress to ALCLs, while others remain healthy. A cocktail of cytokines (IL2, IL7, and IL15) was shown to facilitate low levels of proliferation and efficient lentiviral transduction in the absence of TCR triggering (49). We introduced NPM–ALK, NPM–ALK^Y664F, and NPM–ALK^KD into these cytokine-treated T cells (Fig. 7A) and tracked survival and expansion to determine the NPM–ALK-transformative capacities (Fig. 7B). After 1 month of culture in these cytokines, we observed no T-cell transformation despite seeing high and consistent expression of NPM–ALK and robust T-cell viability. Moreover, we did not observe any difference between T cells that received NPM–ALK or NPM–ALK^Y664F and those that received NPM–ALK^KD, indicating that NPM–ALK expression alone is not sufficient to transform primary human T cells. To determine whether TCR signals are necessary to promote T-cell transformation, we stimulated NPM–ALK-expressing T cells with anti-CD3/28 or anti-CD3 nine days after the initial cytokine-mediated transduction (Fig. 7C and D). In all cases, NPM–ALK- and NPM–ALK^Y664F-expressing T cells transformed, whereas those expressing NPM–ALK^KD did not. in addition, consistent with our earlier findings, T cells expressing the NPM–ALK^Y644F mutant transformed more rapidly than T cells expressing wild-type NPM–ALK after CD3 only or CD3/CD28 stimulation, indicating that costimulation is not necessary for NPM–ALK-mediated transformation of T cells. These ALCLs were indistinguishable from the ALCLs generated by first activating T cells and then transducing based on expansion kinetics, CD30 and NPM–ALK expression (Fig. 7E). It is important to note that these T cells cultured with common gamma-chain cytokines can proliferate, albeit less than CD3/28-activated T cells (Fig. 7F), suggesting that proliferation is not sufficient to enable T-cell transformation, but rather TCR triggered signal transduction is necessary for NPM–ALK-mediated transformation. Because our findings indicate that TCR triggering is an essential part of the NPM–ALK-mediated T-cell transformation process, they suggest that for a T-cell bearing an NPM–ALK gene translocation to gain a malignant phenotype; the T cell must first recognize its cognate pMHC and initiate TCR-dependent signal transduction. This mechanism may provide an explanation for why healthy individuals can harbor T cells containing the highly oncogenic NPM–ALK translocation without developing ALCL.

Studying tumorigenesis of human cells is an emerging and growing field (50). While much can be learned by studying primary tumor samples and established cell lines, this is an endpoint analysis, making it challenging to understand the events driving tumor formation. Viruses such as Epstein–Barr virus, human T-cell lymphotropic virus-1/2, and herpesvirus saimiri can transform lymphocytes, but in these cases, viral proteins play an active role in the transformation process (51). Murine models have been used to uncover some important aspects of malignant cell transformation but overall have not proven to be faithful models of human tumorigenesis (52). In particular, several attempts have been made to generate a murine model of NPM–ALK transformation, but none have resulted in T-cell tumors arising from mature T cells (18, 20). Previously, we found that T cells from human peripheral blood transduced with a lentiviral vector expressing NPM–ALK were able to transform, and such tumors were indistinguishable from patient-derived ALCL when examined in adoptively transferred NSG mice (21).

Here, using the above model, we have investigated the mechanisms required to promote full transformation of NPM–ALK-expressing T cells. We observed regulation of several critical signaling pathways including MYC, mTOR, and the EMT pathway early during transformation, as well as a striking downregulation of T-cell identity and functionality (Figs. 1–3; Supplementary Figs. S2 and S3). Critical T-cell transcription factors including TCF7, BCL11b, and LEF1 were strongly downregulated by NPM–ALK (Fig. 2B). BACH2, which was recently shown to play a critical role in T-cell subset differentiation and production of a memory T-cell phenotype (53) was repressed starting as early as day 6 (Fig. 2B). SATB1, GATA3, and THEMIS, which have important roles in T-cell differentiation and trafficking, were also rapidly downregulated by NPM–ALK (54–56). Simultaneous with the downregulation of mature T-cell–specific genes, we observed an upregulation of a pluripotency signature, including genes such as Sox2, Zeb2, and ETS2 (Fig. 2E), which are normally expressed in embryonic stem cells or early progenitor cells. Thus, by both upregulating programs associated with stemness and downregulating genes associated with T-cell differentiation, the NPM–ALK-induced program leads to malignant transformation.

The above findings are important given that some studies postulate that the ALK⁺ ALCL cell of origin may be a thymic progenitor cell (17–19). The evidence in favor of this scenario includes the methylation of promoters of T-cell–specific genes and a cell morphology that more closely resembles early thymic progenitors (17, 40). In CD4 T cells transformed by NPM–ALK, we observed a downregulation of the same T-cell–specific proteins that display hypermethylation in ALK⁺ ALCL cells (Fig. 2; Supplementary Fig. S3; ref. 17). Moreover, SHP-1 (PTPN6), which is epigenetically silenced in many hematologic malignancies including NPM–ALK⁺ ALCL (57), was downregulated by NPM–ALK as early as day 9 (Fig. 2C and F; Supplementary Fig. S3). The striking similarity in the T-cell–specific genes downregulated by NPM–ALK and those methylated in patient samples further supports the notion that NPM–ALK may promote dedifferentiation of mature CD4 T cells.

We found that, other than disruption of the NPM–ALK catalytic domain (NPM–ALK^KD), no single mutation that impairs tyrosine phosphorylation blocked T-cell transformation (Fig. 4A and B; Supplementary Fig. S4G), suggesting that NPM–ALK recruits a large signaling complex(es), which remains stable when single anchor points are removed. By defining a triple mutant that failed to transform T cells, we uncovered signals that are required to mediate full transformation (Figs. 4F and G and 5; Supplementary Fig. S4F and S4G). We did not observe early repression of MYC signaling, induction of the EMT program, nor sustained STAT3 signaling in NPM–ALK^TrpM when compared with NPM–ALK^KD, whereas NPM–ALK and NPM–ALK^TrpM equally regulated other pathways including glycolysis, TNFα signaling, and apoptosis (Fig. 5B). Importantly, while we observed downregulation of some T-cell–specific factors, we did not observe induction of genes associated with stemness, indicating that early induction of these factors may be important for the dedifferentiation required to facilitate transformation (Fig. 5C). These data suggest that the protein complexes that bind to NPM–ALK, but not NPM–ALK^TrpM (Fig. 5D), modulate some, if not all the pathways, regulated by NPM–ALK that are essential for transformation.

Our data uncovered a set of genes that may favor the expansion and/or survival of particular NPM–ALK clones (Fig. 6). Of note, ARID1A is a tumor suppressor gene frequently disrupted in ovarian cancers (58) and clones with lentiviral integrations in this locus preferentially expanded in five of 29 independent experiments (Fig. 6E and F). ARID1A suppresses HDAC6 activity (59), which suggests that loss of ARID1A activity would result in more HDAC6-mediated epigenetic changes. ARID1A is normally expressed in lymph nodes, and loss of activity has been implicated in genome instability and cancer progression (60, 61), suggesting that ARID1A may be a negative regulator of NPM–ALK transformation. ARID1A downregulation has also been implicated in inducing the EMT program (61), potentially contributing to the transcriptional signal observed here.

Our studies also suggest that a key factor in determining why some individuals that harbor T cells with NPM–ALK translocations develop into ALCL is whether their T cells become exposed to antigens (pMHC). This finding also suggests that a specific TCR-triggered pathway(s), rather than an exit from G₀ or increased metabolic activity, is required to enable NPM–ALK transformation, as cytokine-treated T cells have many of the characteristics of activated T cells, but did not transform (Fig. 7B, E, and F). The cooperation between NPM–ALK and TCR signaling is likely to be essential only during the initial stages of NPM–ALK-mediated transformation because NPM–ALK subsequently downregulates many components of the TCR signaling pathway (17, 40, 62).

In summary, we have developed a robust model of early-stage T-cell oncogenesis that demonstrates the requirement for TCR activation to foster the full transformation of NPM–ALK-expressing T cells. We also identify dynamic regulation of the MYC, EMT, and PI3K/mTOR, signaling pathways early in the transformation process that parallel a downregulation of T-cell identity and function. We further define key residues within NPM–ALK required for binding to signaling complexes that are mandatory to promote and sustain NPM–ALK-mediated oncogenic transformation. Finally, we reveal changes in the gene expression and epigenetic profiles that a mature T cell undergoes to become akin to an immature T cell, making a strong case that ALK⁺ ALCL can originate from mature CD4 T cells. We believe that our findings describing the fundamental mechanisms of NPM–ALK-mediated oncogenesis can serve as a model to better understand factors that promote or repress tumor formation.

F. Wei reports grants from National Natural Science Foundation of China outside the submitted work; in addition, F. Wei has a patent for a method for human T cell immortalization issued. A. Watkins reports employment as a senior scientist at Merck Research Laboratories during conduct of study. F.D. Bushman reports grants from NIH during the conduct of the study. J.L. Riley reports grants and personal fees from Merck, grants and other support from Tmunity Therapeutics, and grants from NIH during the conduct of the study, as well as a patent for US20160194616A1 issued to University of Pennsylvania. No disclosures were reported by the other authors.

J.M. Pawlicki: Conceptualization, funding acquisition, investigation, methodology, writing–original draft, writing–review and editing. D.L. Cookmeyer: Investigation, methodology, writing–review and editing. D. Maseda: Investigation, writing–original draft, writing–review and editing. J.K. Everett: Software, methodology, writing–review and editing. F. Wei: Conceptualization, investigation, writing–review and editing. H. Kong: Investigation, writing–review and editing. Q. Zhang: Investigation, writing–review and editing. H.Y. Wang: Investigation, writing–review and editing. J.W. Tobias: Software, validation, writing–review and editing. D.M. Walter: Investigation, writing–review and editing. K.M. Zullo: Investigation, writing–review and editing. S. Javaid: Investigation, writing–review and editing. A. Watkins: Supervision, writing–review and editing. M.A. Wasik: Supervision, writing–review and editing. F.D. Bushman: Conceptualization, resources, supervision, funding acquisition, writing–original draft, project administration, writing–review and editing. J.L. Riley: Conceptualization, resources, supervision, funding acquisition, writing–original draft, project administration, writing–review and editing.

The authors thank the Penn Center for AIDS Research (P30-AI045008) and Cancer Center Human Immunology Core (P30-CA016520) for providing purified human T cells (Meizan Lai, Kevin Gayout, Chui Yu Lau, and Emily Winters) and performing the immune repertoire profiling experiments (Wenzhao Meng and Eline T. Luning Prak); Frances Male for help with the integration analysis; Penn Bioinformatics Core (Taehyong Kim and Paul Wang) for help with RNA-seq analysis; and the Proteomics & Metabolomics Facility at The Wistar Institute (Hsin Yao Tang and Thomas Beer) for performing the mass spectrometry analysis. They also thank Samik Basu, Raymond Joseph Moniz, Peter Georgiev, Elaine Pinheiro, Sheila Ranganath from Merck; Megan Lim and Kojo Elenitoba-Johnson from the University of Pennsylvania; and the Riley Lab for helpful discussions. The NIH Tetramer Facility is supported by contract HHSN272201300006C from the National Institute of Allergy and Infectious Diseases, a component of the NIH in the Department of Health and Human Services for the IV9 tetramer. Support for these studies was provided by the NIH (T32HL007775 and U19AI117950), Tmunity Therapeutics, and Merck.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.