|Home | About | Journals | Submit | Contact Us | Français|
Diffuse large B-cell lymphoma (DLBCL) comprises disease entities with distinct genetic profiles, including germinal center B-cell (GCB) like and activated B-cell (ABC) like DLBCLs. Major differences between these two subtypes include genetic aberrations leading to constitutive NF-κB activation and interference with terminal B-cell differentiation through BLIMP1 inactivation, observed in ABC- but not GCB-DLBCL. Using conditional gain-of-function and/or loss-of-function mutagenesis in the mouse we show that constitutive activation of the canonical NF-κB pathway cooperates with disruption of BLIMP1 in the development of a lymphoma that resembles human ABC-DLBCL. Our work suggests that both NF-κB signaling, as an oncogenic event, and BLIMP1, as a tumor suppressor, play causal roles in the pathogenesis of ABC-DLBCL.
ABC-DLBCL is the most aggressive DLBCL and has a poor clinical prognosis. Constitutive NF-κB activity interferes with the apoptotic effect of chemotherapy and may account for the poor response to treatment of ABC-DLBCL patients. Our studies in the mouse improve the understanding of human ABC-DLBCL pathogenesis by the demonstration that two recurrent events in this disease: constitutive NF-κB activity and abrogation of terminal B-cell differentiation through BLIMP1 disruption, cooperate in lymphomagenesis. Because of the similarity of the lymphomas arising in the compound mutants with human ABC-DLBCL these mice may serve as a preclinical model for this disease, and be used to identify additional oncogenic events and new therapeutic targets.
Diffuse large B-cell lymphoma (DLBCL) is the most frequent lymphoid malignancy, representing 30 to 40% of all non-Hodgkin lymphomas (Lenz and Staudt, 2010; WHO, 2008). DLBCL comprises disease entities with distinct gene expression signatures and response to therapy. Indeed, studies using gene expression profiling have classified various subtypes of DLBCL according to their putative cell of origin (COO) or consensus clusters (Alizadeh et al., 2000; Monti et al., 2005). In the COO classification, two main subgroups of DLBCL emerged. One is the germinal center B-cell (GCB) like DLBCL, which has a gene expression profile that closely resembles that of normal germinal center (GC) B-cells. The other is activated B-cell (ABC) like DLBCL, with a gene expression profile resembling that of in-vitro activated B-cells (Alizadeh et al., 2000).
DLBCLs carry somatically mutated rearranged immunoglobulin (Ig) V region genes (Lenz and Staudt, 2010; Lossos et al., 2000). Although somatic hypermutation (SHM) of Ig genes may not be entirely GC specific, the GCB-DLBCL gene expression profile in conjunction with often ongoing SHM strongly suggests that this lymphoma is indeed derived from a GC B-cell. In the case of ABC-DLBCL, the cell of origin is less clearly defined and may be either a late GC B-cell, an activated post-GC or even GC unrelated B-cell (Lenz and Staudt, 2010).
A major difference between GCB-DLBCL and ABC-DLBCL is constitutive NF-κB activity in the latter (Alizadeh et al., 2000; Staudt, 2010). NF-κB signaling plays a crucial role in B-cell physiology and can make B-cells independent of survival factors, such as BAFF (Sasaki et al., 2006). Similarly, ABC- but not GCB-DLBCL relies on constitutive activity of the canonical NF-κB pathway for survival (Davis et al., 2001; Staudt, 2010). Recently, mutations leading to constitutive canonical NF-κB activation in ABC-DLBCL have been described (Compagno et al., 2009; Davis et al., 2010; Kato et al., 2009; Lenz et al., 2008a). Another characteristic of ABC-DLBCL are genetic alterations that interfere with terminal B-cell differentiation. Thus, ~25% of ABC-DLBCLs show inactivating mutations of BLIMP1 (Pasqualucci et al., 2006; Tam et al., 2006), a key regulator of plasma cell differentiation (Martins and Calame, 2008), suggesting that BLIMP1 may function as a tumor suppressor in the pathogenesis of ABC-DLBCL. Additional recurrent mutations in ABC-DLBCL that block plasma cell differentiation include genetic aberrations resulting in deregulated expression of SPIB (~26%) or BCL6 (~24%) (Iqbal et al., 2007; Lenz and Staudt, 2010; Lenz et al., 2008b).
In an attempt to assess the roles of NF-κB activation and BLIMP1 disruption in the pathogenesis of ABC-DLBCL we used a genetic system in the mouse that allows conditional gain-of-function and/or loss-of-function mutagenesis in GC B-cells.
For targeted mutagenesis in GC B-cells we used the Cγ1-cre transgene, expressed in B-cells at early stages of the GC reaction (Casola et al., 2006). To induce activation of the NF-κB canonical pathway we combined this transgene with a ROSA26 allele, termed IKK2castopFL, that harbors a cDNA encoding a constitutively active IKK2 protein, which mediates canonical but not alternative NF-κB activation, preceded by a loxP flanked STOP cassette (Sasaki et al., 2006). We complemented this system by introducing a conditional BLIMP1 allele (Blimp1F; (Ohinata et al., 2005)), alone or in combination with IKK2castopFL. Activation of the IKK2castopFL allele by Cre-mediated recombination is marked by expression of GFP under the control of an internal ribosomal entry site downstream of the inserted IKK2 cDNA. To report Cre-mediated recombination in cells carrying the Blimp1F allele, we combined it with a ROSA26 reporter allele harboring an YFP gene preceded by a loxP flanked STOP cassette (eYFPstopFL; (Srinivas et al., 2001)). Mice carrying the Cγ1-cre transgene in combination with eYFPstopFL served as controls.
To determine the impact of enforced NF-κB activation and/or disruption of BLIMP1 on the GC reaction we examined splenic GC B-cells in control and experimental mice upon immunization with sheep red blood cells (SRBC). Mice with BLIMP1 deletion alone showed increased numbers of GC B-cells at day 10 after primary immunization (Figure 1A, C), a trend that persisted to day 21 and upon secondary immunization (Figure 1B, C). In contrast, mice with enforced expression of IKK2ca alone showed some reduction in the GC B-cell fraction at day 10 after primary (Figure 1A, C) and secondary immunization (Figure 1C), to which the increased BLIMP1 levels in these cells may contribute (Figure 1D) (Martins and Calame, 2008). Indeed, concomitant BLIMP1 disruption overrode this effect, and, like deletion of BLIMP1 alone, led to increased numbers of GC B-cells (Figure 1A, C). However, 21 days after primary immunization, the spleens of mice with enforced NF-κB activation alone or together with BLIMP1 loss were virtually devoid of GFPpos GC B-cells (Figure 1B, C). Although constitutive NF-κB activation did not affect AICDA transcript levels in GC B-cells 10 days after immunization, the few remaining GC B-cells in these mice on day 21 post-immunization carried a significantly lower number of somatic mutations in their Ig heavy chain (IgH) V regions compared to controls (Figure 1E, F). Collectively these results suggest that constitutive NF-κB activity promotes premature termination of the GC reaction through both BLIMP1 dependent and independent mechanisms, thereby reducing the load of somatic mutations in the GC progeny.
With respect to cell proliferation, BLIMP1 deletion in GC B-cells, with or without concomitant activation of NF-κB, resulted in a small, but significant increase of the fraction of cells in the S and G2 phases of the cell cycle, compared to controls (Figure 2A). This could be due to the loss of BLIMP1 (Figure 1D), which is thought to repress genes promoting cell cycle progression, in the small fraction (0.5 to 4%) of BLIMP1pos GC B-cells (Angelin-Duclos et al., 2000; Lin et al., 1997; Martins and Calame, 2008). In accord with the notion of an anti-proliferative effect of BLIMP1, its acute ablation also promoted the proliferation of B-cells in an in-vitro cell culture system that mimics T-cell dependent B-cell activation (Figure 2B-D). In this in-vitro cell culture system, the efficiency of Cre-mediated recombination was identical in B-cells from mice of the various genotypes, and virtually complete by day 3 of culture, as measured by the expression of the reporter genes (Figure 2B).
Interestingly, enforced IKK2 expression in the in-vitro system not only inhibited apoptosis of the activated B-cells (Figure 2E), but also enhanced their proliferation and cooperated with BLIMP1 loss in accumulating increased numbers of cells over the 5-day culture period (Figure 2C, F). These results contrast with both the absence of an effect of enforced IKK2ca expression on the cell cycle distribution of GC B-cells in-vivo (Figure 2A) and the IKK2ca mediated premature termination of the GC reaction (Figure 1). We considered the possibility that constitutive NF-κB signaling in GC B-cells might interfere with BCL6 expression (Saito et al., 2007), but could not detect such an effect when we analyzed BCL6 mRNA expression by quantitative PCR (data not shown).
As expected from the literature (Martins and Calame, 2008), ablation of BLIMP1 in B-cells through Cγ1-cre severely impeded plasma cell formation in both Cγ1-cre;Blimp1FFeYFPstopFL and Cγ1-cre;Blimp1FFIKK2castopFL mice, 10 days after primary and secondary immunization (Figures 3A, B; and S1A, B), accompanied by a strong reduction of total as well as SRBC specific IgG1, but not IgM, antibodies in the sera of the animals (Figures 3C; S1C; and data not shown). Thus, in agreement with BLIMP1’s essential role in terminal B-cell differentiation, its inactivation in our experimental system blocks plasma cell differentiation, also when combined with constitutive NF-κB activation. In contrast, activation of NF-κB alone was compatible with plasma cell differentiation (Figures 3A, B; and S1A, B), and antibody titers in the sera of Cγ1-cre;IKK2castopFL mice 10 days after primary immunization were similar to those in control mice (Figures 3C; and S1C).
To assess the role of constitutive activation of the canonical NF-κB pathway and/or disruption of BLIMP1 in lymphomagenesis, we generated cohorts of compound mutant mice and monitored them for tumor formation over a period of 550 days. Mice with enforced expression of IKK2ca alone showed a similar life span as control mice (Figure 4). In contrast, deletion of BLIMP1 in GC B-cells led to shortened mouse survival, a phenotype further enhanced by the constitutive activation of the canonical NF-κB pathway (Figure 4).
Macroscopic examination of Cγ1-cre;IKK2castopFL mice, sacrificed between 550 to 600 days of age, revealed the presence of enlarged spleens in all cases (6/6), with a significant hyperplasia of both B- and plasma cells in spleen and bone marrow (Figure 5A, B; and S2A, B). Accordingly, large numbers of cells expressing intracellular Ig and the plasma cell marker CD138 could be detected in splenic histological sections of these mice, whereas there was essentially no expression of the GC B-cell marker PNA (Figure 5C; and data not shown). Consistent with the increased numbers of plasma cells, the serum titers of both IgM and IgG1 were higher in Cγ1-cre;IKK2castopFL mice than controls (Figure S2C). Electrophoretic analysis of sera from these mice revealed the presence of distinct band/s in the γ globulin region of the gel (M-spikes) in 70% of Cγ1-cre;IKK2castopFL mice (7/10) compared to 20% (2/10) of controls, indicative of clonal or oligoclonal plasma cell expansions (Figure 5D; and data not shown). Thus, constitutive activation of the NF-κB canonical pathway promotes not only B-cell hyperplasia, consistent with our previous results (Sasaki et al., 2006), but also plasma cell hyperplasia.
Examination of terminally ill Cγ1-cre;Blimp1FFeYFPstopFL mice revealed splenomegaly in all cases (11/11), occasionally together with lymphadenopathy, and hepatomegaly in one case (Figure 6A; and data not show). Histological analysis of enlarged spleens and lymph nodes showed in 6/7 mice examined the presence of large cells with a diffuse growth pattern, resembling human DLBCL (Figure 6B). These cells were immunohistochemically negative for CD138 and intracellular Ig, but expressed the B-cell marker CD19 and thus represented transformed B rather than plasma cells (Figure 6C; and data not shown). The analysis of IgH gene rearrangements by Southern blot in spleen and lymph nodes revealed that the lymphoproliferations were of clonal origin in 6/8 cases (Figure 6D). We succeeded in amplifying IgH rearrangements from 3 clonal tumors by PCR and found that two of them were highly somatically mutated, while the third one carried a single nucleotide exchange (Table S1). We then assessed whether these lymphomas resembled a particular DLBCL subtype. Expression of MUM1/IRF4 as identified by immunohistochemical staining correlates with ABC-, while BCL6 expression is primarily associated with GCB-DLBCL (Choi et al., 2009; Hans et al., 2004), although a fraction of ABC-DLBCLs is also positive for BCL6 (Alizadeh et al., 2000; Iqbal et al., 2007; Lenz and Staudt, 2010). Consistent with an ABC-DLBCL phenotype, 5/6 DLBCLs arising in Cγ1-cre;Blimp1FFeYFPstopFL mice were IRF4posBCL6neg by immunohistochemical staining while the remaining case was IRF4negBCL6pos and in addition stained for the GC B-cell marker PNA (Figures 6E; and S3A, B). Since IRF4 is a target of the NF-κB pathway (Grumont and Gerondakis, 2000; Saito et al., 2007), and most human ABC-DLBCLs show NF-κB activation we looked whether this pathway was activated in lymphomas arising upon BLIMP1 loss. Using the YFP reporter as a marker, we purified lymphoma cells from the IRF4negBCL6pos (L1) and two IRF4posBCL6neg (L2, L3) lymphomas, for which we had material available, and determined the transcript levels of known NF-κB target genes frequently expressed in human ABC-DLBCL (Alizadeh et al., 2000). Consistent with a GCB-DLBCL phenotype, the L1 lymphoma had low expression of NF-κB target genes and elevated transcript levels of GC B-cell markers (Figure 6F). In contrast, the L2 and L3 lymphomas showed elevated expression of some (NFKBIA, PIM2, BCL2) or all NF-κB target genes tested, but low transcript levels of GC B-cell markers (Figure 6F). We searched for mutations of the A20, CARD11 and CD79a and b genes in these cells, which have been associated with constitutive NF-κB activation in human ABC-DLBCL. However, no such mutations were detected by sequence analysis of amplified cDNAs. We also tested the expression levels of non NF-κB target genes typically expressed in human ABC-DLBCL (Wright et al., 2003), including the transcription factor FOXP1, indicative of a poor clinical prognosis (Banham et al., 2005; Lenz and Staudt, 2010). Both L2 and L3 had elevated levels of FOXP1 compared to GC B-cells, however only L3 showed expression of the majority of the other ABC-DLBCL associated genes tested (Figure 6F). Together, our results demonstrate that BLIMP1 inactivation promotes lymphoma development in the mouse, in agreement with the recent results of others (Mandelbaum et al., 2010). In addition, the lymphomas arising in these mice often show molecular features of human ABC-DLBCL.
Mice with combined BLIMP1 deletion and activation of the NF-κB pathway had a significantly reduced life span compared to mice with BLIMP1 inactivation alone (Figure 4). Similar to the latter, these mice succumbed to a B-cell derived lymphoproliferative disease affecting spleen and lymph nodes, showing in all 5 cases analyzed morphological features resembling human DLBCL (Figure 6A-C; and data not shown). Analysis of IgH rearrangements by Southern blot showed that these mice develop clonal lymphomas (Figure 6D). Sequence analysis the of clonal IgH rearrangements amplified by PCR revealed the presence of several somatic point mutations in one case and a single nucleotide exchange in another, while no mutations could be detected in two other cases. These latter tumors, like the one with a single mutation, expressed IgM on their surface (Table S1). Consistent with an ABC-DLBCL phenotype and constitutive NF-κB activation, 5/5 lymphomas analyzed were IRF4posBCL6neg by immunohistochemical staining (Figures 6E; and S3A). Analysis of purified (GFPpos) tumor cells from three of these lymphomas for which we had material available (L4, L5, L6), revealed low expression of GC B-cell markers and elevated transcript levels for most of the ABC-DLBCL associated NF-κB target and non-target genes tested, compared to GC B-cells (Figure 6F). Collectively, these results reveal a strong cooperative effect between constitutive activation of the canonical NF-κB pathway and BLIMP1 loss in the pathogenesis of lymphomas resembling human ABC-DLBCL.
Using conditional mutagenesis in GC B-cells in the mouse we show that BLIMP1 disruption promotes the development of lymphomas often resembling human ABC-DLBCL. This is in full accord with the results of Mandelbaum et al (Mandelbaum et al., 2010). The low penetrance, long latency and clonality of DLBCL in these mice suggested that additional oncogenic events are required for the pathogenesis of this disease. A major candidate for one such event is the NF-κB pathway, since the majority of human ABC-DLBCLs display constitutive NF-κB activation, and most lymphomas arising upon conditional BLIMP1 deletion in mice also show activation of this pathway. Indeed, we found that constitutive activation of the canonical NF-κB pathway strongly synergized with loss of BLIMP1 in lymphomagenesis. Thus our work suggests that both NF-κB signaling and the loss of BLIMP1 play causal roles in the pathogenesis of human ABC-DLBCL.
Similar to ABC-DLBCL, genetic alterations leading to constitutive NF-κB activation have been described in multiple myeloma, suggesting a role for this pathway in the pathogenesis of this disease (Annunziata et al., 2007; Keats et al., 2007; Staudt, 2010). Supporting this concept, we found that constitutive activation of NF-κB through the Cγ1-cre transgene promotes plasma cell hyperplasia. If ABC-DLBCLs indeed derive from an activated B-cell at an early stage of plasma cell differentiation (Lenz and Staudt, 2010), these results in turn indicate that the pathogenesis of this disease necessarily requires interference with terminal B-cell differentiation, as exemplified by BLIMP1 inactivation.
Besides abrogating B-cell terminal differentiation, the loss of BLIMP1 may promote cellular proliferation through derepression of BLIMP1 target genes controlling cell cycle progression (Lin et al., 1997; Martins and Calame, 2008). Constitutive activation of the canonical NF-κB pathway, on the other hand, releases B-cells from their dependence of extracellular survival signals, such as BAFF, and although by itself not promoting cell division, enhances the proliferative response of mitogen stimulated B-cells (Sasaki et al., 2006). Thus, the cooperative effect of BLIMP1 disruption and constitutive NF-κB signaling in the pathogenesis of human ABC-DLBCL likely results from blocking the differentiation of an activated B-cell that displays deregulated proliferative and survival properties.
The increased and abnormally prolonged GC reaction in Cγ1-cre;Blimp1FFeYFPstopFL mice in conjunction with the presence of somatically mutated Ig gene rearrangements in the lymphomas arising upon BLIMP1 deletion, suggests that the latter likely originate predominantly from either a late or a post-GC B-cell. The situation is less clear in the case of BLIMP1 ablation in combination with activation of the NF-κB pathway, where we did not detect somatic mutations in the majority of the tumors. It is possible that this reflects an inherent feature of our experimental approach, in that NF-κB activation early on in the GC reaction leads to its premature termination ((Bolduc et al., 2010; Kishi et al., 2010) and the present work). It may become significant in this situation that the Cγ1-cre transgene induces Cre-mediated recombination also in a small fraction of IgM expressing non-GC B-cells, whose absolute numbers easily compare to those of GC B-cells (Casola et al., 2006). Such cells likely include antigen-activated B-cells that have not yet entered the GC reaction (Casola et al., 2006; Garside et al., 1998; Toellner et al., 1996), and to whose proliferative expansion and ultimate transformation Cre-mediated NF-κB activation and loss of BLIMP1 would equally contribute as in the case of B-cells activated in the course of the GC reaction. Indeed, the lymphomas arising in the compound mutants and carrying unmutated Ig gene rearrangements expressed IgM on the cell surface. In this scenario, the present mouse model reproduces the pathogenesis of a B-cell lymphoma originating from the transformation of an activated B-cell, and thus of human ABC-DLBCL. In the latter case, however, the activated cell of origin usually represents a B-cell in which the SHM mechanism is or has been operating, because of its increased risk to accumulate unwanted mutational events (Lenz and Staudt, 2010). This feature applies less stringently in our system of conditional mutagenesis.
The lymphomas arising upon disruption of BLIMP1 alone and in combination with constitutive NF-κB activation in compound mutant mice are of clonal origin, suggesting that additional oncogenic event/s occurred which led to the outgrowth of a particular B-cell clone. Because of the similarity of these lymphomas with human ABC-DLBCL, the identification of these additional genetic lesions might lead to the discovery of critical players in ABC-DLBCL pathogenesis and open the way to new therapeutic strategies.
Cγ1-cre, IKK2castopFL, Blimp1FF and eYFPstopFL alleles have been described (Casola et al., 2006; Ohinata et al., 2005; Sasaki et al., 2006; Srinivas et al., 2001). Eight- to 10-week-old mice were immunized i.v. with 1×109 SRBCs (Cedarlane) in PBS. Mouse cohorts were monitored twice a week for tumor development under monthly antigenic stimulation by SRBC immunization, and euthanized if signs of tumor development occurred. All animal care and procedures followed National Institutes of Health guidelines and were approved by the Institutional Animal Care and Use Committee (IACUC 03341) of Harvard University and the Immune Disease Institute.
Lymphoid single-cell suspensions were stained with the following monoclonal antibodies from BD, αCD19(ID3), αB220(RA3-6B2), αCD95(Jo2), αCD138(281-2), αCD38(90). Samples were acquired on a FACSCantoII (BD), and analyzed using FlowJo software (Tree star).
Elisa was performed as described (Casola et al., 2006). For detection of SRBC specific IgG1, 2×106 SRBCs were incubated with sera for 20min at 4°c, washed twice with 1×PBS and then incubated with αmouseIgG1 (A85-1, BD) antibody for 20min at 4°c. After one wash with 1×PBS, the geometric mean of the fluorescence intensity was determined through the acquisition of SRBCs samples by flow cytometry.
Serum was diluted 1:2 in barbital buffer and analyzed on a Hydragel K20 system according to manufacturer’s instruction (Sebia).
Total RNA was extracted using TRIzol reagent and cDNA was synthesized using Thermoscript RT-PCR system (Invitrogen). For qRT-PCR, we used Power SYBR Green, followed by analysis with StepOnePlus system (Applied Biosystems). Samples were assayed in duplicate and messenger abundance was normalized to that of HPRT. To detect BLIMP1 transcripts we used a primer set pairing in the exon 4-5 junction. Since in the Blimp1F allele exon 5 is loxp flanked and will be deleted upon Cre-mediated recombination, (Ohinata et al., 2005)), we used this primer set to confirm BLIMP1 deletion. Heat map was generated using the matrix2png software (Pavlidis and Noble, 2003).
This was done by Southern blotting of EcoRI digested genomic DNA from tumors, using a JH probe spanning the JH4 exon and part of the downstream intronic sequence.
Tissues were fixed with 10% formalin (Sigma) and paraffin embedded sections were stained with hematoxylin and eosin (H&E, Sigma), αIRF4 (MUM1, Santa Cruz), αBCL6 (sc-C19, Santa Cruz), PNA (Vector), αCD138 (281-2, BD), and αmIg (Vector).
Cells were stained for surface markers and fixed with 2% formaldehyde for 15 min at RT and then incubated with the DNA-binding agent Draq5 (Biostatus) at a final concentration of 12.5 μM for 30min at RT. Samples were acquired and analyzed as above. Cell cycle analysis was performed using the Watson-Pragmatic computational model in Flow-Jo software (Tree star).
Genomic DNA was prepared from tumor tissues or sorted B-cells. IgH-V gene rearrangements were PCR-amplified using the Expand High fidelity PCR system (Roche) and forward primers VHA and VHE (Ehlich et al., 1994), and a reverse primer in the JH4 intron (5’-CTCCACCAGACCTCTCTAGACAGC-3’). Fragments were cloned, sequenced and blasted against the NCBI database (http://www.ncbi.nlm.nih.gov/igblast/) to determine VHDHJH usage. The cloned intronic sequences were then aligned to their germline counterparts. In determining somatic mutations, we excluded polymorphisms associated with the Cγ1-cre allele.
Splenic B-cells were purified by CD43 depletion (Miltenyi). Cells were cultured in the presence of 1 μg.ml-1 of αCD40 (HM40-3, eBioscience) and 25ng.ml-1 of IL-4 (R&D Systems). To monitor cellular division, B-cells were labeled in 1ml of 2.5μM CFSE (Invitrogen) in 1×PBS per 107 cells at 37°C for 10 min. Before acquisition of CFSE labeled cells by flow-cytometry, inactivation of GFP or YFP fluorescence was achieved by fixing cells using a Fix/Perm reagent (eBioscience). Cells undergoing apoptosis were detected using Active Caspase-3 Apoptosis Kit (BD), according to manufacturer’s instructions.
Data were analyzed using unpaired two-tailed student’s t-test, a p value ≤0.05 was considered significant. * p≤0.05, ** p≤0.01, *** p≤0.001, (ns) non significant. Survival curves were compared using the Logrank test. Data in text and figures is represented as mean ±(Standard Error of the Mean) SEM.
We thank D. Ghitza, A. Pellerin, J. Grundy and J. Xia for technical assistance, Rajewsky lab members, M. Janz and S.A. Godinho for critical comments and suggestions, S. Koralov for help with sequence analysis and S. Peng for HPRT qRT-PCR primers. K.R. is supported by the National Cancer Institute through P01CA092625 and an LLS SCOR grant, and D.P.C and B.Z. by postdoctoral fellowships of the Leukemia & Lymphoma Society.
The authors declare no conflict of interest.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.