Transformation of mouse T cells requires MYC and AKT activity in conjunction with inhibition of intrinsic apoptosis

Peripheral T-cell lymphoma is an aggressive non-Hodgkin's lymphoma characterized by excessive proliferation of transformed mature T cells. The number and nature of genetic aberrations required and sufficient for transformation of normal T cells into lymphomas is unknown. Here, using a combinatorial in vitro-approach, we demonstrate that overexpression of MYC together with activated AKT in conditions of inhibition of intrinsic apoptosis rapidly resulted in transformation of mature mouse T cells with a frequency approaching 100%. Injection of transformed cells into mice resulted in rapid development of aggressive T cell lymphoma, characterized by spread to several organs, destruction of tissue architecture and rapid death of the animals. TcR-sequencing revealed a polyclonal repertoire of tumor cells indicating that co-expression of MYC, activated AKT and BCLXL is sufficient for tumor transformation and do not require acquisition of additional genetic events. When analyzing cells with inducible expression we found that proliferation of transformed T cells required sustained expression of both MYC and AKT. AKT exerted a dual function as it inhibited induction of, and promoted exit from, cellular quiescence and contributed to inhibion of apoptosis. Downregulation of AKT and/or MYC together with BCLXL resulted in rapid and complete elimination of cells through induction of apoptotic cell death.


INTRODUCTION
Peripheral T-cell lymphomas (PTCLs) or mature T-cell lymphomas are a group of usually aggressive non-Hodgkin lymphomas (NHLs) accounting for around 10-15 % of NHL cases [1][2][3]. The disease is associated with a poor outcome and there is need for improved treatment regimens. Development of new treatments should be facilitated by an improved knowledge about the molecular mechanisms underlying the disease. PTCLs are highly heterogeneous, consisting of several subgroups. The three most common subgroups, PTCL not otherwise specified (PTCL-NOS), angioimmunoblastic T-cell lymphoma (AITCL) and anaplastic large cell lymphoma (ALCL) account for around 60% of the cases. ALCL can further be divided into two subsets based on presence or absence of recurrent translocations involving the anaplastic lymphoma kinase (ALK) at 2p23 [4]. A subset of PTCL-NOS carry a recurrent translocation t(5:9)(q33;q22) creating a fusion between the ITK and SYK genes resulting in the expression of the ITK-SYK fusion protein [5]. Recently, several recurrent mutations have been identified in AITCL as well as in a subset of PTCL-NOS with an AITCL-associated gene expression profile. These include mutations in the GTPase RHOA [6,7] as well as in the epigenetic regulators TET2 [8,9], DNMT3A [9], and IDH2 [10]. Whereas mutations in RHOA appear specific for AITCL, mutations in TET2, DNMT3A and IDH2 are also common in myeloid malignancies [11,12]. Mutations in a number of genes including PLCG1, CD28 [13], FYN [7], CTNNB1 (beta-catenin) [13], STAT3, JAK1 and TP53 [14] as well as translocations involving IRF4 [15] have also been identified in PTCL. Furthermore, deregulated expression of a large number of genes associated with other hematopoietic tumors as well as other cancers have been identified in PTCL, including SNF5 [16], LIN28B [17], MYC [18,19], PI3K [18], mTOR [18], AKT [19,20], MAF [21], genes involved in Notch signaling [22], members of the Polycomb repressive complex 2 including BMI1 [23] as well as genes regulating intrinsic [24] and extrinsic apoptosis [25].
Collectively, these studies indicate that although several disease-associated genes may contribute to the development of PTCL-like disease, the prolonged time preceding tumor development and the monoclonality of resulting tumors (where analyzed) in these experimental models indicate that additional, yet unidentified, genetic events were required for tumor development. Also, the notion that mature T cells may be resistant to oncogene driven transformation has been put forward [39]. An important, and so far, unanswered question is therefore if normal mature T cells can be tumor transformed, and in that case what would be the number and nature of driver events required. Herein, as a step towards an increased understanding of the cellular and molecular requirements for transformation, starting from a combinatorial in vitroassay for T cell transformation, combinations of effector driver-genes sufficient for rapid, high frequency tumor transformation of mature murine T cells were identified.

Overexpression of distinct gene combinations results in T cell transformation
Primary murine T cells were transduced at day 2 and 3 after anti-CD3 stimulation with a pool of 12 individual retroviruses encoding genes reported mutated or having altered expression patterns in PTCL and other tumors, namely MYC, BCLXL, BMI1, IRF4, dominant negative mutant p53 (p53DD), constitutively active myristoylated AKT (Myr-AKT), constitutively active β-catenin, constitutively active Notch (ICN1), a constitutively activated form of STAT3 (STAT3c) and a myristoylated constitutively active PI3 kinase (Myr-PIK3CA) as well as activated HRAS (HRAS-V12) and hTERT, followed by culture in the absence of anti-CD3 ( Figure 1A-1B). As an initial criterion for potential transformation, expansive growth exceeding four weeks was chosen. Transduction of T cells with this pool of retroviruses reproducibly resulted in transformation ( Figure 1B). Subsequently, one individual gene at a time was eliminated, followed by T cell transduction. Only in the pools lacking either MYC or Myr-AKT (AKT), did transformation not occur, indicating that MYC and AKT were both indispensable for transformation ( Figure 1B). We subsequently tested pools of retroviruses containing MYC and AKT plus one additional construct. Four distinct combinations of genes leading to transformation were identified namely; MYC+AKT+BCLXL (reproduced > 15 times with cells from different mice and independent viral preparation), MYC+AKT+BMI1 (reproduced > 5 times), MYC+AKT+IRF4 (reproduced > 5 times) and MYC+AKT+p53DD (reproduced > 5 times) ( Figure 1C). We tested if over-expression of other apoptotic inhibitors than BCLXL, including one additional members of the BCL2-family, MCL1, IAP-family members, cIAP2 and XIAP, inhibitors of the death receptor-mediated pathway of apoptosis, FLIP L and FLIP S as well as the dominantnegative mutants FADD-DN and RIP-DN, could cooperate with MYC and AKT in inducing T cell transformation, which was not the case ( Figure 1D). It should be noted that absence of effects of some genes or gene combinations tested herein do not exclude their eventual importance during T-cell transformation but could reflect limitations in experimental design.

Co-expression of MYC, AKT and BCLXL leads to rapid and high frequency-transformation of mature T cells
As bicistronic vectors were used, MYC, AKT and BCLXL expression could be monitored through YFP, GFP and DsRed-monomer expression by flow cytometry (Figure 2A). Co-transduction of cells with MYC, AKT and BCLXL rapidly resulted in exponential growth, whereas expression of one or two of the genes in combination did not ( Figure 2B). In cultures co-transduced with the three genes, cells co-expressing MYC, AKT and BCLXL rapidly outgrew other cells ( Figure 2C). Limiting dilution assays were performed at day three after retroviral transduction. About 1 in 15 of total viable cells formed expandable clones ( Figure 2D). Considering the number of input T cells co-expressing MYC, AKT and BCLXL (20%, Figure 2D), the corrected frequency of T cell transformation was 1 in 3 (1/2 -1/4, 95% confidence limits). Transformation by co-expression of MYC, AKT www.oncotarget.com and BCLXL was observed in T cells from three out of three tested mouse strains (C57BL/6, BALB/c and 129/ Sv) ( Figure 2E). Plate-bound anti-CD3, exposure to ConA or allo-stimulation all resulted in transformation by coexpression of MYC, AKT and BCLXL (Supplementary Figure 1A), indicating that the mode of T cell activation did not seem to affect permissiveness for transformation.
Surface phenotypes of the transformed cells revealed a large degree of heterogeneity concerning expression of the T cell markers CD4, CD8, CD3 and CD90. Often, cells were double positive (DP) for CD4 and CD8 with variable expression of CD3. (Supplementary Figure 1A-B). The phenotypes of clones obtained in limiting dilution also showed variable phenotypes ranging from CD4 and CD8 single positive (SP) cells, DP and double negative (DN) cells (Supplementary Figure 2). To directly test if SP cells were targets of transformation, T cells were sorted into CD4 and CD8 SP cells (> 99% purity, Figure  2F) followed by transformation, resulting in rapid entry into exponential growth ( Figure 2F). Phenotypic analyses clearly indicated that transformed cells started off as SP T cells but thereafter lost or gained expression of CD4, CD8 (Supplementary Figure 3), reminiscent of the variable expression of T-cell associated surface-receptors in human PTCL [40].

Transformed T cells form aggressive tumors in vivo
In an initial experiment, injection of mice with T17 cells (MYC+AKT+BCLXL) resulted in death of 3 out of 4 animals already 12 days after injection, whereas mice injected with A20 cells (MYC+AKT+p53DD) were judged moribund at days 28-31. All mice injected with A20 cells had at least one enlarged kidney and flow cytometric analyzes revealed large infiltrations of tumor cells in kidneys but also to some extent in livers ( Supplementary  Figures 4, 5, Supplementary Table 1). Mice injected with independently derived T17 cells showed signs of aggressive disease and were considered moribund 9 days after injection and were euthanized ( Figure 3A). The mice exhibited splenomegaly and massive numbers of tumor cells were detected in the bone marrow, spleen, liver and thymus ( Figure   mice were stimulated with plate bound anti-CD3 in vitro for two days, followed by transduction with combinations of retroviruses encoding (grey) or not (white) the indicated genes. Cultures were scored positive where growth could be recorded, through visual inspection, for more than 4 weeks. www.oncotarget.com tumor cells in the kidneys but not in the thymus ( Figure  3B, Supplementary Table 2B). Mice injected with T20 cells (MYC+AKT+p53DD) from an independent experiment, were euthanized between days 20-40, displayed no signs of lymphadenopathy and splenomegaly was observed only in one out of five mice. Tumor cells were hardly

Expression of MYC, AKT and BCLXL is sufficient for tumor transformation of mature T cells
In the experiments above, cells were kept in culture during 8-12 weeks, prior to injection into mice. A possibility would be that rare cells, with potential additional genetic lesions required for tumor growth, could have been selected for during in vitro culture or after injection in vivo. As genetic aberrations are rare events, it would be expected that if additional genetics events, in conjunction with expression of MYC, AKT and BCLXL would be required for tumor growth in vivo, then this would result in monoclonal tumors, and conversely, if not, tumors would be polyclonal. T cells were transduced and around 3.5x10 5 cells co-expressing markers for MYC, AKT and BCLCL (Supplementary Figure 10) were injected into mice already three days after transduction. Mice were euthanized 8 days after injection and bone marrow, spleen and liver contained very high levels of infiltrating tumor cells co-expressing the fluorescent reporters for MYC, AKT and BCLXL (Supplementary Table 3). Reporter-expressing T cells accounted for more than 90% of all T cells in all analyzed organs (Supplementary Table 3). Genomic DNA was prepared from spleens of three diseased animals and the T cell receptor beta variable (TcR-Vβ) genes were sequenced (Supplementary Table 4 and Dataset 1). The TcR-Vβ gene usage was similar between all sequenced samples giving an indication of a broad repertoire of tumor transformed T cells in the spleens of the three diseased mice (Supplementary Table 5). Around 4000 unique productive TcR-Vβ gene rearrangements were identified in each spleen from the diseased animals (Supplementary Dataset 1). When subtracted for the number of TcR-Vβ gene rearrangements possibly corresponding to endogenous T cells, assuming these would have the smallest clone sizes, around 3500 unique productive TcR-Vβ gene rearrangements per mouse spleen were identified at this relatively low level of sequencing depth ( Figure 4B). This indicates that the repertoire of tumor transformed T cells in this experimental setting is polyclonal in turn suggesting that genetic aberrations in addition to overexpression of MYC, AKT and BCLXL would not be required for tumor transformation of mature T cells.
However, a clear bias towards larger clone sizes could be observed in the spleens of the diseased mice as compared to normal spleen or anti-CD3 stimulated splenic T cells ( Figure 4A), indicating clonal competition among tumor-transformed T cells. We therefore compared frequencies of 686 identical clones that were identified in all three mice ( Figure 4B). Although the overall clone size distribution of the shared clones was similar ( Figure  4C) individual clone sizes differed considerably among the three mice ( Figure 4D, Supplementary Dataset 1), reflected in low Pearson correlation coefficients ( Figure  4E), indicating that in this experimental setting, retroviral integration affecting either the expression levels of the transgenes or possibly affecting expression or function of neighboring genes, or the specificity of the TcR, did not to any larger extent influence the degree of clonal expansion of tumor-transformed T cells in vivo.

Dependence of MYC, AKT and BCLXL expression for sustained proliferation and survival in vitro
In order to gain insight into the mechanistic roles of MYC, AKT and BCLXL in transformed cells, T cells with inducible expression ("Tet-on") of MYC, AKT and BCLXL, or combinations thereof, were derived. Using real-time PCR we verified that MYC, AKT and BCLXL expression was downregulated as a consequence of doxycycline (Dox) washout ( Figure 5A). Downregulation of MYC or AKT arrested cellular expansion, resulted in rapid accumulation of cells in the G 1/0 phase of the cell cycle and a reduction in cell size, although this occurred with some delay after downregulation of AKT as compared to MYC. MYC downregulation was not associated with apoptotic cell death whereas apoptosis was detected after AKT downregulation, compatible with an anti-apoptotic effect of AKT [41]. ( Figure 5B). BCLXL downregulation, as expected, led to induction of apoptotic cell death, paralleled by loss of cell viability but the around 60% fraction of apoptotic cells remained rather constant, from day 2 up till day 7. Despite the initial reduction in total numbers of live cells, between days 4 to 7, cell numbers increased exponentially to a similar rate as in cultures expressing BCLXL ( Figure 5B). BCLXL downregulation resulted in a decreased proportion of (non-apoptotic) cells in G 1/0 phase at least between days 1 to 4 ( Figure  5B), in accordance with previous reports (reviewed in [42]). The reasons for the lower dependence on BCLXL expression in established cell lines as compared to during the initial transformation event (Figures 1-2) is unknown. Simultaneous downregulation of both MYC and AKT resulted in similar consequences as downregulation of MYC or AKT individually. Downregulation of MYC or AKT, together with downregulation of BCLXL however, resulted in massive apoptotic cell death with few, if any, live cells remaining 3 days after Dox washout. Similar data were obtained when simultaneously down-regulating all three genes ( Figure 5B).
Subsequently we investigated whether downregulation of MYC and/or AKT had resulted in induction of a reversible or irreversible state of cell cycle arrest, ie, quiescence or senescence. After 3 days in the absence of Dox after which the cells had entered a state of cell cycle arrest, Dox was re-added, which resulted in in re-induction of exponential growth, suggesting that downregulation of MYC and/or AKT resulted in induction of cellular quiescence ( Figure 6A). Reinduction of BCLXL resulted in decreased frequency of apoptotic cells, increased viability although not affecting, the rate of cellular expansion ( Figure 6A). We then analyzed consequences of prolonged periods of AKT downregulation (7, 13 and 30 days) before re-induction. As expected, cells rapidly accumulated in G 1/0 accompanied by a gradual increase in the fraction of apoptotic cells, reaching 89%, 38 days after Dox washout. At all time points, re-induction of AKT resulted in reversal of cell cycle arrest, re-induction of exponential proliferation, increased cell size and decreased apoptotic rate ( Figure  6B). Thus, our data would suggest that AKT in addition to counteracting quiescence induction also partially inhibited apoptotic cell death.

DISCUSSION
Our results suggest that expression of MYC and activated AKT together with inhibition of intrinsic apoptosis via expression of BCLXL would constitute sufficient conditions for tumor transformation of mature mouse T cells, based on the rapid kinetics and high frequency of the process. This conclusion was further supported by the polyclonal repertoire of transformed T cells identified in vivo after injection into mice. Transformed T cells in this experimental setting could be considered malignant as they rapidly spread to different organs, resulted in destruction of normal tissue organization and leading to early death of the animals. To our knowledge, this is the first description of defined genetic factors sufficient for transformation of mature T cells. Our results do not support the notions that mature T cells would be resistant to oncogene transformation [39] unless occurring in a monoclonal T cell environment [43]. However, differences in experimental design as well as the genes tested, LMO2, TCL1, a constitutively active mutant of TrkA and NPL/ALK fusion protein, compared to those assayed herein, likely can explain the different results obtained.
As expression of BMI1 or dominant negative p53 could substitute for BCLXL, the main tumor suppressive effect of p53 that had to be inhibited in order for transformation to occur in mature T cells, likely consisted in inhibition of intrinsic apoptosis. In addition to promoting proliferation, pro-apoptotic effects of MYC are well documented and can at least partly be inhibited by expression of anti-apoptotic Bcl-2 family proteins reviewed in [44]. In accordance, co-expression of MYC together with BCLXL accelerated tumorigenesis in several experimental models [45,46]. From these data, it could be anticipated that overexpression of MYC and BCLXL would be required and maybe sufficient, for T cell transformation. However, although we found that overexpression of MYC and BCLXL in mature T cells was required, transformation did not occur. Instead, we unexpectedly identified a concomitant requirement for expression of activated AKT in order for transformation to occur. Mechanistically, we found that AKT exerted a dual function as it inhibited induction of, and promoted exit from, cellular quiescence and also to some extent inhibited induction of intrinsic apoptosis, in accordance with previous reports [41,47]. In established tumor cell lines, downregulation of AKT resulted in cellular quiescence accompanied by a rather modest rate of apoptotic cell death, up to 15% live cells remained after 35 days. Simultaneous downregulation of both AKT and BCLXL however, resulted in complete cellular eradication via apoptotic cell death within days. If these results could be extrapolated to human T-cell tumors (and maybe also other tumor types), then an effective treatment regimen, if tolerated, could consist in simultaneous inhibition of AKT or TcR-signaling and inhibition of anti-apoptotic Bcl-2 proteins. In line with this, synergy between inhibition of members of the anti-apoptotic Bcl-2 family and inhibition of PI3K/AKT/mTor signaling has recently been reported for human myeloid leukemia and diffuse large B-cell lymphoma cells [48,49]. Elucidation of the exact molecular mechanisms underlying the interplay of MYC and AKT during T cell transformation, would warrant further studies and the cells described herein, with inducible expression of AKT, would be suitable for such an exploration. It is however challenging as MYC may be able to regulate up to a third of all genes and numerous AKT substrates have been described [50,51].
Differences in requirements for tumor transformation of mouse and human cells have been described and it is therefore an open question if human T cells would be permissive for transformation with the same gene combinations as we here describe for mouse T cells [52]. Nevertheless, we believe our findings are of clinical relevance in PTCL as a number of mutated genes, including RHOA, CD28, LCK, FYN, PDK1, PIC3CA, PIK3R1, AKT1 [6,7,13,53], as well as the fusion proteins NPM-ALK and ITK-SYK [4,5,26,54,55] either are part of, or can activate, the TcR/PI3K/ AKT signaling pathway and our prediction is therefore that expression of at least some of these mutated genes would have the ability to substitute for activated AKT in the experimental system described herein. Moreover, deregulated expression of MYC [18,19], members of the Polycomb repressive complex 2 including BMI1 [23] as well as genes regulating intrinsic apoptosis [24] has been reported in PTCL. These genes we would denote as "effector drivers", able to directly affect cellular expansion as opposed to "diversity drivers", genes mutated or with altered expression that contribute to genetic or epigenetic instability or alterations in splicing fidelity, exemplified by TET2, DNMT3A and IDH2, also commonly mutated in PTCL [8][9][10]13], thereby generating a diverse and abnormal repertoire of cells with altered gene functions and/or expression levels, allowing for selection and expansion of a subset of cells with altered expression of one or several effector driver genes.
Despite the aggressive characteristics of PTCL, it is difficult to establish cell lines from PTCL-patients, reflected in the modest number of cell lines available [56]. This could indicate that in many cases, PTCL cells would not (yet) have acquired a state of autonomous growth, in contrast to the tumor-transformed T cells describe herein. An interesting possibility would be that some (epi)-genetic events leading to MYC or AKT activity or inhibition of p53 activity and/or intrinsic apoptosis in PTCL cells would be substituted for by receptor/ligand interactions taking place in the tumor microenvironment in vivo.
It is our belief that the description of sufficient conditions for tumor transformation of T cells provided herein as well as the experimental model for T cell transformation will provide a framework for further dissection of the mechanisms behind tumor transformation of hematopoietic cells in general and for PTCL in particular.

Stimulation of T cells and retroviral transduction
Retroviral particles were produced as described followed by concentration using through centrifugation at 6000 × g overnight at +4°C [57]. Splenic cells (1×10 6 /ml) were stimulated with plate-bound anti-CD3 antibody (clone 2C11), 2.5 μg/ml ConA (GE Health Care) or irradiated (1500Rad) allogeneic BALB/c splenic cells in complete RPMI medium [58]. Where indicated, T cells were sorted with a BD FACSAria sorter (BD Biosciences, San Jose, CA) into CD4 + and CD8 + cell populations, followed by stimulation with plate-bound anti-CD3 antibody in the presence of soluble anti-CD28 antibody. Retroviral transduction of around 5x10 5 stimulated spleen cells was performed at day 2 and day 3 after stimulation using spin infection in the presence of 8 μg/ml polybrene (Sigma-Aldrich). Cells were replated in culture medium without stimuli at day 6 and growth was followed for more than 4 weeks. For cloning of cells using limiting dilution, 60 cultures for each dilution with syngeneic irradiated (2000 rad) thymocytes (2.5×10 6 /ml) as feeder cells were used. Clones were thereafter expanded to substantial numbers in the absence of feeder cells, followed by phenotypic analysis. Results were analyzed as described [59].

Flow cytometry
Cells from in vitro cultures or single cell suspensions obtained from organs from euthanized mice were stained with mAbs: anti-mouse CD4, anti-mouse CD8, antimouse CD3 and anti-mouse CD90 (BD Biosciences Pharmingen, San Diego, CA). Propidium iodide was used to evaluate viability and to eliminate dead cells from phenotypic analyses. Mean fluorescence intensity in the forward scatter channel (FSC) was used to estimate relative cell size. Cell cycle analysis was performed as described [73]. Cells were analyzed on a BD LSRFortessa Cell Analyzer (BD Biosciences) and the data analyzed with FlowJo software (Tree Star, Inc., San Carlos, CA). EGFP and EYFP signals were separated with a 526 nm dichroic mirror and 510/20 nm and 550/30 nm bandpass filters. Where indicated, numbers of viable cells were determined by cell counting using a BD LSRFortessa Cell Analyzer connected to a BD High Throughput Sampler (BD Biosciences). The number of live cells in 100 μl of cell suspension was determined in triplicates.

RNA isolation and real-time PCR
Total RNA was isolated followed by synthesis of cDNA. Real-time PCR was carried out using TaqMan gene expression pre-synthesized reagents directed against MYC, AKT and BCLXL and Gapdh using a master mix in a 7500 Real-Time PCR system (Applied Biosystems, Foster City, CA). Relative gene expression was calculated using the ΔΔCt method.

High throughput TcR V-beta deep sequencing
Genomic DNA was prepared using DNEasy minicolumns (Qiagen). Sequencing was performed www.oncotarget.com by Adaptive Biotechnologies (Seattle, WA) using the Immunoseq platform and data were analyzed using the ImmunoSEQ Analyser software (Adaptive biotechnologies).

Author contributions
KH, SD, TPF and AG performed the experiments. AG designed the study and wrote the manuscript with input from KH and SD.