RelB-Dependent Stromal Cells Promote T-Cell Leukemogenesis

Background The Rel/NF-κB transcription factors are often activated in solid or hematological malignancies. In most cases, NF-κB activation is found in malignant cells and results from activation of the canonical NF-κB pathway, leading to RelA and/or c-Rel activation. Recently, NF-κB activity in inflammatory cells infiltrating solid tumors has been shown to contribute to solid tumor initiation and progression. Noncanonical NF-κB activation, which leads to RelB activation, has also been reported in breast carcinoma, prostate cancer, and lymphoid leukemia. Methodology/Principal Findings Here we report a novel role for RelB in stromal cells that promote T-cell leukemogenesis. RelB deficiency delayed leukemia onset in the TEL-JAK2 transgenic mouse model of human T acute lymphoblastic leukemia. Bone marrow chimeric mouse experiments showed that RelB is not required in the hematopoietic compartment. In contrast, RelB plays a role in radio-resistant stromal cells to accelerate leukemia onset and increase disease severity. Conclusions/Significance The present results are the first to uncover a role for RelB in the crosstalk between non-hematopoietic stromal cells and leukemic cells. Thus, besides its previously reported role intrinsic to specific cancer cells, the noncanonical NF-κB pathway may also play a pro-oncogenic role in cancer microenvironmental cells.


Introduction
The Rel/NF-kB transcription factors function in multiple biological processes, including development, immunity, inflammation, and response to cellular stress [1]. NF-kB subunits are often activated in solid or hematological malignancies as the result of rearrangements/mutations in their genes or in genes encoding components of the NF-kB signaling pathway, persistent autocrine or paracrine stimulation through specific cell surface receptors, or viral or cellular oncoprotein activity (for review see [2,3]). NF-kB activation in cancer cells has been shown to activate genes involved in cell survival, proliferation, angiogenesis, invasion, and chemoresistance being therefore an important target for cancer therapy. Recently, an important function for the canonical NF-kB pathway in inflammatory cells infiltrating several types of solid tumors has been brought to light. NF-kB activation in those cells leads to the production of cytokines, growth factors, and angiogenic factors that promote malignant conversion and progression (for review see [3]).
The NF-kB proteins are transcriptional regulators that bind cognate DNA elements as homo-or heterodimers. NF-kB activity is controlled by interaction with IkB (inhibitor of NF-kB) proteins and only when these are degraded by the proteasome, following serine phosphorylation by IkB kinases (IKK) and ubiquitination, are NF-kB dimers released. The NF-kB/Rel family comprises five members (RelA, RelB, c-Rel, p50/p105, and p52/p100) sharing the conserved Rel homology domain, which is responsible for DNA binding, nuclear localization, dimerization, and IkB binding. In contrast to RelA (p65), RelB, and c-Rel, the p50 and p52 proteins, which derive from proteolytic processing of the p105 and p100 precursor proteins, respectively, lack transactivation domains. The p50 and p52 proteins act thus as transcriptional repressors, except when forming heterodimers with other NF-kB members or when interacting with other transcriptional activators, such as the Bcl3 protein (for review see [1]).
Two main NF-kB activation pathways have been identified [1]. The canonical NF-kB activation pathway, which is triggered by an array of stimuli such as proinflammatory cytokines, antigen receptors, Toll-like receptors, and cellular stress, relies on IKKb (IKK2)/IKKc (NEMO)-dependent IkB phosphorylation and degradation and results in RelA and/or c-Rel activation. Disruption of the canonical pathway in immune cells impairs innate and acquired immune responses in a cell-autonomous or non cell-autonomous manner (for review see [4]). The noncanonical NF-kB activation pathway, which can be activated by specific members of the TNF receptor family (e.g., lymphotoxin b receptor [LTbR] and BAFF receptor [BAFF-R]) depends on IKKa (IKK1) and NIK kinase activity but not on IKKb or IKKc [1]. Upon stimulation, IKKa phosphorylates p100 on C-terminal serine residues and induces its ubiquitin-dependent processing to generate p52. When released from p100 sequestration, p52:RelB, p50:RelB, and, as recently shown, p50:RelA dimers shuttle to the nucleus to activate transcription of specific target genes [5][6][7][8]. Disruption of the noncanonical pathway also affects immune cell function, impairing either lymphoid organogenesis due, at least in part, to defective LTbR signaling, or mature B cell function and maintenance due to defective BAFF-R signaling [9]. Furthermore, inactivation of the noncanonical pathway breaks down central tolerance as a result of impaired generation of medullary thymic epithelial cells (mTEC), which are essential for negative selection of autoreactive T cells [9][10][11].
Most studies in human lymphoid leukemia and lymphoma have identified canonical NF-kB activation in leukemic cells. For example, NF-kB activation is frequently observed in Hodgkin's lymphomas due to activation of the CD30, CD40, and RANK receptors or due to inactivating mutations in the IkBa-encoding gene [12]. Activation of p50 homodimers and p50:RelA heterodimers was detected in all major subtypes of human acute lymphoblastic leukemia (ALL) [13].
The v-rel oncogene, the retroviral counterpart of c-rel, induces aggressive leukemia/lymphoma in chicken and transgenic mice [14]. Canonical NF-kB activity was also found in T-ALL induced in mice following expression of a Tal1 transgene or of intracellular Notch1 (Notch1-IC) oncogenic protein [15,16]. Finally, both the canonical and noncanonical NF-kB pathways were found to be activated by viral oncoproteins, in particular the HTLV1-encoded Tax protein in adult T-cell leukemia and the EBV-encoded LMP1 protein in B-cell lymphoma [12].
Several reports indicate that the noncanonical NF-kB pathway is also activated in specific subtypes of lymphoid leukemia and lymphoma (for review see [9,12,17]). Chromosomal translocations disrupting the Nfkb2 gene that generate truncated p100 proteins and constitutive processing of p100 to p52 were identified in cutaneous T-cell lymphoma, and, more rarely, in B-cell non-Hodgkin lymphoma, chronic lymphocytic leukemia, and multiple myeloma [9,12,17]. Transgenic expression of a truncated p100 protein led to the development of B-cell lymphomas in mice, thus demonstrating the oncogenic potential of Nfkb2 mutations [18]. Recently, genetic alterations in components of the noncanonical and canonical NF-kB pathways (e.g., NIK, TRAF3, CYLD, BIRC2/ BIRC3, CD40, NFKB1, or NFKB2) resulting in their activation have been identified in multiple myeloma [19,20].
In the present report we assessed the role of NF-kB proteins in a transgenic mouse model for human T-ALL induced by the TEL-JAK2 fusion protein [21,22] and uncovered a specific role for RelB in T-cell leukemia development. Using RelB knockout mice we found that RelB assisted TEL-JAK2-induced T-cell leukemogenesis. Interestingly, bone marrow chimeric mouse experiments showed that RelB is not required in the hematopoietic compartment but plays a role in radio-resistant stromal cells to favor leukemia onset and increase disease severity.

TEL-JAK2 leukemic cells display RelA and RelB activation
To assess the NF-kB activity status in leukemic T cells from EmSRa-TEL-JAK2 transgenic mice, we performed electrophoretic mobility shift assays (EMSA) on nuclear extracts using an NF-kB-specific oligonucleotide probe. Nuclear extracts from TEL-JAK2 tumor cells showed significantly higher levels of NF-kB DNAbinding activity, visible as two bands with different mobility, as compared to control thymocyte nuclear extracts ( Figure 1A). To determine which NF-kB members were activated in TEL-JAK2 leukemic cells, we pre-incubated nuclear extracts with specific NF-kB antibodies that either supershift or inhibit protein/DNA complexes. A p50/NF-kB1 antibody quantitatively supershifted band I and most of band II in TEL-JAK2 leukemic cells ( Figure 1B, compare lanes 1 and 2). The remaining NF-kB activity in band II was inhibited by a p52/NF-kB2 antibody ( Figure 1B, lane 3), indicating the presence of both proteins in different complexes. Band II complexes also included RelA and RelB since the RelA antibody inhibited a slowly migrating complex within band II ( Figure 1C, lane 2 and Figure S1A To identify the NF-kB dimers activated in TEL-JAK2 leukemic cells, we performed supershift analyses combining different antibodies. Combination of RelA and p50 antibodies yielded the Figure 1. Activation of NF-kB transcription factors in TEL-JAK2 leukemic cells. (A) Nuclear extracts obtained either from freshly isolated wild-type control thymocytes (WT thy) or from thymus (thy) or lymph node (LN) leukemic cells from two representative TEL-JAK2 (TJ2) diseased mice (nu66 and nu13) were analyzed by electrophoretic mobility shift assays (EMSA) using the NF-kB-binding Ig-kB oligonucleotide probe. Similar results were obtained with more than 20 primary leukemic samples. Bands labeled I and II (arrowheads) indicate specific NF-kB complexes as assessed by competitive EMSA (data not shown). An Ets-specific probe (E74) was used for nuclear extract normalization. (B-D) Antibody supershift analysis of Ig-kB-bound complexes using the indicated antibodies was performed on leukemic nuclear extracts from individual TEL-JAK2 tumors (tumor nu1284 in panel B and D; tumor nu38 in panel C). Open arrows indicate supershifted complexes, filled arrows identify the different NF-kB complexes, and the asterisk indicates a nonspecific band. p50, p52, RelA, and RelB DNA-binding activity in TEL-JAK2 leukemic cells was also detected using a palindromic NF-kB probe derived from the IL2Ra promoter, indicating that the previous results were not biased by differential DNA recognition by NF-kB proteins. doi:10.1371/journal.pone.0002555.g001 same migration pattern as that observed with p50 antibody alone ( Figure 1D, compare lanes 2 and 5), indicating that RelA heterodimerized with p50. Combination of the RelB antibody with p50 and RelA antibodies inhibited the remaining complex ( Figure 1D, compare lanes 5 and 6), which, given that RelB is unable to form homodimers [23], corresponds to p52:RelB heterodimers. Likewise, combination of the RelB antibody with p52 and RelA antibodies inhibited a complex that corresponded to p50:RelB heterodimers ( Figure 1D, compare lanes 7 and 8). In sum, supershift analyses discerned the presence of p50:p50 homodimers as well as p50:RelA, p50:RelB, and p52:RelB heterodimers in TEL-JAK2 leukemic cells.

Nfkb1 deficiency does not affect TEL-JAK2-induced T-cell leukemia development
To assess the role of NF-kB proteins in TEL-JAK2-induced leukemogenesis, we bred EmSRa-TEL-JAK2 mice with mice deficient for specific NF-kB genes. To prevent p50-containing complex formation, we first bred EmSRa-TEL-JAK2 mice with Nfkb1 knock-out mice, which do not express the NF-kB1 proteins p50 and its precursor p105 and do not show any thymocyte maturation defects [24]. EmSRa-TEL-JAK2;Nfkb1 2/2 and EmSRa-TEL-JAK2;Nfkb1 +/2 littermate mice developed T-cell leukemia with full penetrance and similar latency ( Figure 2A; median survival of 12 weeks). In addition, Nfkb1-deficient leukemic cells presented a cell surface marker phenotype (expression of variable levels of CD4, CD8, CD24, CD25, and TCRb/CD3e) similar to that of Nfkb1-proficient cells (data not shown) and characteristic of transgenic TEL-JAK2 leukemia [21,25].
EMSA analyses showed that the TEL-JAK2;Nfkb1 2/2 leukemic cells displayed severely reduced NF-kB activity and, as expected, did not display any p50 DNA-binding activity, as evidenced by the lack of p50 homodimers (band I) ( Figure 2B). Interestingly, no RelA DNA-binding activity was detectable in nuclear extracts from TEL-JAK2;Nfkb1 2/2 leukemic cells ( Figure 2C). This was not due to an intrinsic inability to activate RelA in these cells, since TEL-JAK2;Nfkb1 2/2 leukemic cells induced p52:RelA and RelA:RelA DNA-bound dimers upon in vitro treatment with PMA plus ionomycin ( Figure S2). This result suggests that activation of p50 and/or RelA does not play a nonredundant role in TEL-JAK2 leukemogenesis. The only DNA/protein complexes identified in TEL-JAK2;Nfkb1 2/2 leukemic cells were p52:RelB heterodimers, since formation of this complex was inhibited by antibodies against either p52 or RelB, but not by p50-, c-Rel-, or RelA-specific antibodies ( Figure 2C and data not shown).

Generation of viable Relb-deficient mice
Since TEL-JAK2;Nfkb1 2/2 leukemic cells only presented constitutive p52:RelB activity, we set out to evaluate the role of RelB in TEL-JAK2-induced leukemogenesis by generating Relbdeficient TEL-JAK2 transgenic mice. Relb-deficient mice present fatal T-cell-dependent multiorgan inflammation, a phenotype resulting from the absence of mTEC, which are essential for negative selection of autoreactive T cells [26][27][28]. To generate viable Relb-deficient mice, these were bred with Tcra-deficient mice, which do not express the ab T-cell receptor (TCR) and therefore lack mature T cells. The resultant Tcra 2/2 ;Relb 2/2 mice were born at expected Mendelian ratios and lived without external signs of inflammation or other abnormal phenotype, with the exception that female mice failed to nurse their pups. Histological analysis showed no inflammatory infiltrates in liver, lungs, and skin of Tcra 2/2 ;Relb 2/2 mice ( Figure S3A and data not shown), demonstrating that Tcra deficiency rescued the inflammatory phenotype observed in Relb-deficient mice [26][27][28]. In line with previous findings [28], we observed that mature T-cell deficiency in Relb-null mice prevented splenomegaly and splenic myeloid hyperplasia (data not shown).

Relb deficiency delays the onset of TEL-JAK2-induced Tcell leukemia
Our previous studies have shown that breeding of TEL-JAK2 transgenic mice on a Tcra-deficient background does not delay leukemia onset and incidence [25], although TEL-JAK2;Tcra 2/2 leukemic cells showed reduced RelA DNA-binding activity as compared to Tcra-proficient leukemic cells while maintaining a similar level of RelB DNA binding activity ( Figure S5). In contrast, when EmSRa-TEL-JAK2 mice were bred on a Tcra 2/2 ;Relb 2/2 background, we found that these mice developed T-cell leukemia with statistically significant delayed onset as compared to TEL-JAK2;Tcra 2/2 ;Relb +/+ littermates (median survival of 18.5 and 13 weeks, respectively; P,0.01) ( Figure 4A). Diseased mice from both groups presented leukemic cells in thymus, spleen, lymph nodes, bone marrow, liver, and lungs ( Figure S6 and data not shown). However, TEL-JAK2;Tcra 2/2 ;Relb 2/2 mice presented significantly reduced tumor load in thymus and lymph nodes, as compared to Relb-proficient littermates ( Figure 4B). Similar to Relb-proficient cells, Relb-deficient leukemic cells presented the variable levels of CD4, CD8, CD24, and CD25 cell surface markers that characterize TEL-JAK2 leukemic cells ( Figure S7). TEL-JAK2;Tcra 2/2 ;Relb 2/2 leukemic cells showed similar p50:p50 and p50:RelA NF-kB DNA-binding activity as TEL-JAK2;Tcra 2/2 ;Relb +/2 leukemic cells (data not shown), indicating that RelB inactivation did not lead to selection of leukemic cells displaying enhanced DNA-binding activity of other NF-kB family members. Together, these results reveal a non-redundant function for RelB in TEL-JAK2-induced leukemogenesis. However, because RelB deficiency affects all mouse tissues, these results do not distinguish whether RelB function is required intrinsically in the hematopoietic cells targeted by TEL-JAK2, and/or whether it is required in non-leukemic cells from the tumor microenvironment to support TEL-JAK2-induced leukemogenesis.

Relb deficiency in the hematopoietic compartment does not affect TEL-JAK2-induced T-cell leukemogenesis
To analyze whether the function of RelB in TEL-JAK2-induced leukemogenesis reflected an intrinsic role in the hematopoietic compartment, we generated bone marrow radiation chimeric mice. Bone marrow cells obtained from either TEL-JAK2;Tcra 2/2 ; Relb 2/2 or TEL-JAK2;Tcra 2/2 ;Relb +/2 non-diseased mice were intravenously injected into wild-type (WT), lethally irradiated recipient mice ( Figure 5A). Both groups of bone marrow chimeric mice developed T-cell leukemia with similar onset (median survival of 18 and 17 weeks for TEL-JAK2;Tcra 2/2 ;Relb +/2 RWT and TEL-JAK2;Tcra 2/2 ;Relb 2/2 RWT mice, respectively) ( Figure 5B). Diseased chimeric mice of both groups developed leukemia affecting the thymus, frequently in association with pleural effusion, lymph nodes, bone marrow, and occasionally spleen and liver (data not shown). No significant difference in lymphoid organ invasion and leukemic cell surface marker expression was observed between the two experimental groups ( Figure S8 and data not shown). Of note, under these experimental conditions non-TEL-JAK2 transgenic Tcra 2/2 ;Relb 2/2 and Tcra 2/2 ;Relb +/2 bone marrow donor cells were both capable to normally reconstitute the T-cell compartment when adoptively transferred into lethally irradiated syngeneic WT mice, as demonstrated by the similar proportions of DN and DP thymocytes and by the absence of mature CD4 or CD8 SP cells ( Figure S9). Together, these results indicate that Relb deficiency in the hematopoietic compartment hampered neither normal T-cell development nor TEL-JAK2-induced T-cell leukemogenesis.

Discussion
In the present study we have found that expression of the RelB transcription factor in non-hematopoietic, stromal cells promotes T-cell leukemogenesis induced by the TEL-JAK2 oncoprotein. These conclusions stem from two sets of data showing delayed leukemia onset in (i) TEL-JAK2 transgenic mice deficient in RelB, as compared to RelB-proficient littermates, and (ii) lethally irradiated RelB-deficient mice reconstituted with TEL-JAK2 transgenic bone marrow, as compared to similarly reconstituted RelB-proficient littermates. These results thus demonstrate that, in addition to their reported pro-oncogenic role intrinsic to leukemic cells in several mouse models [16,34] and in human leukemia cell lines [12], NF-kB transcription factors contribute to T-cell leukemogenesis through the generation and/or maintenance of a proper stromal microenvironment.
Relb 2/2 mice develop lethal multiorgan inflammation shortly after birth [26,27]. This disease is caused by the appearance of autoreactive mature T cells resulting from the absence of mTECs expressing Aire transcriptional regulator-dependent peripheral tissue antigens, which are important to establish central selftolerance [26][27][28]35]. The possibility that the delay in TEL-JAK2induced leukemogenesis observed in RelB-deficient mice could be linked to the inflammatory phenotype of these mice can be excluded since our experiments were performed in the Tcra 2/2 genetic background, which prevents the development of autore-active T cells. Indeed, in contrast to the reported RelB-deficient phenotype [26][27][28], no significant inflammatory infiltrates were observed in the organs of Tcra 2/2 ;Relb 2/2 mice.
Our previous results showed that TEL-JAK2 transforms immature DN and DP thymocytes [25]. Since DN and DP thymocyte development remained unaffected in the Tcra 2/2 ; Relb 2/2 thymic microenvironment, we exclude the possibility that the delay in TEL-JAK2-induced leukemogenesis in RelB-deficient mice was simply due to a reduction in cellular targets available for oncogenic transformation. Our results thus suggest that the RelBdependent microenvironment contributes specifically to DN/DP thymocyte transformation by TEL-JAK2.
Although lymph node development is strongly affected by the lack of RelB, the direct cellular defects associated with RelB deficiency in this organ are as yet unknown. In contrast, it has been shown that RelB-deficient thymi lack a defined medulla and mTECs and show a strong reduction in CD80 + DEC205 + dendritic cell (DC) numbers secondary to the defect in thymic architecture and mTECs [26,27,36]. Accordingly, Tcra 2/2 ;Relb 2/2 mice showed no discernable thymic medulla and no UEA-1 + mTECs. TCRa-deficient mice also show thymic architectural defects, with reduced and disorganized medulla and fewer UEA-1 + mTECs than wild-type mice [11,31,32]. These defects are restored by adoptive transfer of mature T cells [11]. The combined deficiency of Tcra and Relb, but not Tcra deficiency alone [25], delayed TEL-JAK2-induced leukemia onset, thus indicating that, contrary to the RelB-deficient thymic defects, those found in TCRa-deficient mice have no detectable impact on leukemia development. Gray et al [31] have recently shown that TCRa-deficient thymi lack MHC II lo /Ly51 2 (mTEC lo ) cells, while Kaplan-Meier leukemia-free survival curves for Tcra 2/2 recipient mice deficient or not for the Relb gene that received bone marrow from EmSRa-TEL-JAK2;Tcra 2/2 transgenic mice were significantly different (log-rank test, P value,0.01). A mouse deceased from an unrelated cause was censored in the analysis (tick mark). (C) Thymus and lymph node weights were plotted for TEL-JAK2;Tcra 2/2 RTcra 2/2 ;Relb 2/2 and TEL-JAK2;Tcra 2/2 RTcra 2/2 ;Relb +/+ diseased chimeric mice. *, P value,0.05; ***, P value,0.001 (unpaired t-test). The number of mice analyzed in each group is given between parentheses. doi:10.1371/journal.pone.0002555.g006 RelB-deficient thymi additionally lack MHC II hi /Ly51 2 (mTEC hi ) cells. It is thus tempting to speculate that specifically mTEC hi cells assist TEL-JAK2-induced T-cell leukemogenesis, although we cannot exclude an additional requirement for a RelB-dependent function in other stromal cells including DCs or cortical thymic epithelial cells. Moreover, RelB-dependent thymic stromal cells may assist TEL-JAK2 leukemogenesis either directly, through cell-cell contact or paracrine growth factor stimulation, or indirectly by stimulating other stromal cells (e.g., DCs) to interact with leukemic cells.
The nature of the molecular signals emanating from the thymic or lymph node stroma that favor T-cell leukemia initiation or progression remains to be identified. It is likely that RelB activity in mTEC or lymphoid organ stromal cells induces the expression of genes that favor T-cell leukemogenesis. Proteins known to play a role in thymic function include cytokine/growth factors (e.g., IL-7, Kit ligand, Notch ligands, and Sonic Hedgehog), chemokines (e.g., Ccl19/Elc, Ccl21/Slc, Ccl25/Teck, and Cxcl12/Sdf-1), cell surface receptors (e.g., LTbR, RANK, and Notch-1 to -3) and adhesion molecules (e.g., ICAM-1, MAdCAM-1, and VCAM-1) [37,38]. RelB DNA-binding activity can be stimulated by RANK and LTbR, two receptors coupled to NF-kB activation and shown to be important for thymic medulla and lymphoid organ formation [39]. Both receptors activate NF-kB through the canonical and noncanonical pathways, with RANK specifically requiring TRAF6. LTbR signaling in thymic mTECs and in lymph node DCs induces expression of Ccl19 and Ccl21 [40,41], which are known RelB target genes [42,43], and of these chemokines as well as MAdCAM1, ICAM-1, and VCAM-1 in lymph node stromal cell organizers [44]. Since TEL-JAK2 leukemic cells express the Ccr7 transcript, encoding the receptor for the Ccl19 and Ccl21 chemokines (data not shown), and display cell surface expression of the ICAM-1 receptor LFA-1 (data not shown), it is tempting to speculate that these NF-kB signaling-dependent targets may play a role in TEL-JAK2-induced leukemogenesis.
Recent studies have shown that the composition of the thymic stroma is dynamic and modulated by particular stimuli (e.g., LTab, FGF family members, Wnt, and steroids) [45]. It is thus possible that leukemic T cells analogously induce qualitative and/ or quantitative changes in thymic stromal populations. Our transcriptomic analysis showed higher expression levels of the LTa-and LTb-encoding genes in TEL-JAK2 leukemic cells as compared to normal thymocytes ( [25] and data not shown). It is therefore possible that LTa 1 b 2 production by leukemic cells may modulate the thymic microenvironment in its favor through interaction with LTbR-expressing stromal cells and in this way contribute to leukemogenesis.
Our data cannot discriminate whether RelB-dependent stromal cells facilitate the initiation or the progression of T-cell leukemia, or both. Nevertheless, the limited tumor burden in thymus and lymph nodes of terminally ill TEL-JAK2;Tcra 2/2 ;Relb 2/2 and TEL-JAK2;Tcra 2/2 RTcra 2/2 ;Relb 2/2 mice suggests that the RelB-dependent thymic microenvironment favors the expansion of transformed leukemic cells. During normal T-cell development, Tcra 2/2 ;Relb 2/2 thymi presented a slight reduction in thymocyte cellularity as compared to Tcra 2/2 ;Relb +/2 thymi, suggesting that both normal and leukemic T cell expansion is affected by RelBdependent thymic stroma. The thymic medulla is a compartment where IL-7-dependent proliferation of mature T cells occurs before their export to the periphery [46,47]. This compartment in RelB-proficient TEL-JAK2 mice could favor the expansion of leukemic cells in the thymus.
Stromal cells from lymphoid organs or from the bone marrow are important to sustain survival and proliferation of human leukemic cells. The survival of T-ALL primary cells in vitro is promoted by exogenous growth factors (e.g., IL-7) or by co-culture with bone marrow or thymic stromal cells [48][49][50][51]. Also, thymectomy was shown to prevent T-cell leukemia development induced by Ikaros deficiency in mice [52], further supporting an important role of the thymic microenvironment in T-cell leukemia development. That tumor microenvironmental-derived signals are required for the in vivo expansion of TEL-JAK2-induced leukemic cells is supported by the fact that these cells survived for over a week ex vivo but failed to proliferate under these conditions (data not shown).
TEL-JAK2 leukemic cells displayed NF-kB activity (p50, p52, RelA, and RelB), suggesting that, in addition to a noncellautonomous role, NF-kB activation may also contribute cellautonomously to TEL-JAK2 leukemia development or maintenance. Intrinsic canonical NF-kB activity has recently been shown to be important for Notch-induced murine T-cell leukemia, since disease development was inhibited by expression of the IkBa super-repressor mutant [16,34]. This effect could be specific to these particular models of T-cell leukemia, since the IkBa superrepressor failed to affect Tal-1-induced T-ALL [15]. Transgenic co-expression of the IkBa super-repressor with TEL-JAK2 neither did inhibit NF-kB activity nor affected leukemia incidence or severity (N.d.S., Marie Körner, and J.G., unpublished data). Likewise, Nfkb1 deficiency failed to affect TEL-JAK2-induced leukemogenesis. Interestingly, TEL-JAK2;Nfkb1 2/2 leukemic cells did not present any RelA DNA-binding activity. In addition, RelA DNA-binding activity was reduced in TEL-JAK2;Tcra 2/2 compared to TEL-JAK2;Tcra +/2 leukemic cells. These observations together with the fact that TCRa-deficiency did not affect TEL-JAK2 leukemia onset [25], indicates that RelA DNA-binding activity correlates neither with leukemia time of onset nor with disease progression. In addition, bone marrow adoptive transfer experiments showed that RelB deficiency in hematopoietic cells did not affect TEL-JAK2-induced leukemogenesis. These results indicate that cell-autonomous RelB expression is not essential for TEL-JAK2 leukemia initiation or maintenance. However, since NF-kB functional redundancy may exist among the different members of the NF-kB family and since complete NF-kB activation could not be experimentally achieved in TEL-JAK2 leukemic cells, we cannot rule out that NF-kB plays a cellautonomous role, in addition to a non-cell-autonomous role, in TEL-JAK2-induced leukemogenesis.
The p52 and RelB proteins are activated by the noncanonical NF-kB activation pathway, which depends on NIK/IKKainduced p100 processing. Accumulating evidence shows that the noncanonical NF-kB pathway is activated in the tumor cells of several human cancers, including B-and T-cell leukemia/ lymphoma, mammary carcinoma, and prostate cancer [12,17,19,20,43,53,54]. The present results are however the first to uncover a role for RelB in the crosstalk between stromal and leukemic cells. Thus, the noncanonical NF-kB pathway may play a pro-oncogenic role both in tumor cells and in cells that compose the tumor microenvironment. In other settings, this property is shared by the canonical NF-kB activity which also plays an important role in activating both intrinsic genetic programs in cancer cells and the production of pro-oncogenic growth factors in cancer-associated stromal or inflammatory cells [3]. It will thus be important to understand how the thymic or lymphoid organ microenvironment assists T-cell leukemia development and to identify the leukemia-promoting factors that are dependent on noncanonical NF-kB activity in stromal cells. A better understanding of the crosstalk between leukemic T cells and stromal cells may pave the way for the development of new therapeutic strategies targeting this disease.

Mice
The EmSRa-TEL-JAK2 transgenic mice (line 71) [21] were bred with Tcra (obtained from CNRS-CDTA; Orléans, France), Relb [27], and Nfkb1 [24] knock-out mice on the C57BL/6 background. All mice were maintained under specific-pathogen-free conditions in the animal facility of the Institut Curie (Orsay, France). All experimental procedures were performed in strict accordance with the recommendations of the European Community (86/609/EEC) and the French National Committee (87/848) for the care and use of laboratory animals. All animal experiments were carried out under the supervision of J.G., who was authorized by the director of the Veterinary Services of the Préfecture de l'Essonne (agreement number 91-7). TEL-JAK2 transgenic mice were euthanized when terminally ill, due to either severe dyspnea caused by massive expansion of leukemic cells in the thymus or extreme weakness caused by leukemic dissemination to vital organs such as bone marrow, lung, and liver. Statistical analyses and survival curves were calculated using Prism 4 (GraphPad, San Diego, CA). Genomic PCR of the sex chromosome-localized Jarid1c/Kdm5c and Jarid1d/Kdm5d genes was performed as described by Clapcote and Roder [55].

Cell isolation and culture
Single cell suspensions were prepared from lymphoid organs gently dissociated and filtered through 70-mm cell strainers. To determine thymocyte cellularity, Trypan Blue (Sigma-Aldrich, St. Louis, MO)-negative viable cells were counted on KOVA microscope slides (Hycor Biomedical, Garden Grove, CA). For in vitro stimulation, cells were cultured at 37uC at 5610 6 cells/ml in RPMI 1640 (Invitrogen, Carlsbad, CA), 5% heat-inactivated FBS (Invitrogen), 50 mM b-mercaptoethanol (Sigma-Aldrich) and stimulated with PMA and ionomycin (both from Sigma-Aldrich) at the indicated concentrations for the indicated period of time.

Flow cytometry
Single cell suspensions from lymphoid organs were stained with fluorochrome-labeled antibodies and detected by a FACSCalibur cytometer (BD Biosciences, San Jose, CA) as previously described [21].