|Home | About | Journals | Submit | Contact Us | Français|
Chronic myeloid leukemia (CML) is induced by BCR-ABL1 and can be effectively treated for many years with Imatinib until leukemia cells acquire drug resistance through BCR-ABL1 mutations and progress into fatal B lymphoid blast crisis (LBC). Despite its clinical significance, the mechanism of progression into LBC is unknown. Here we show that LBC but not CML cells express the B cell-specific mutator enzyme AID. We demonstrate that AID expression in CML cells promotes overall genetic instability by hypermutation of tumor suppressor and DNA repair genes. Importantly, our data uncover a causative role of AID activity in the acquisition of BCR-ABL1 mutations leading to Imatinib-resistance, thus providing a rationale for the rapid development of drug resistance and blast crisis progression.
Activation-induced cytidine deaminase (AID) promotes somatic hypermutation and class switch recombination of immunoglobulin (Ig) genes in germinal center (GC) B cells (Muramatsu et al., 2000; Revy et al., 2000). AID initiates both processes by deaminating Ig gene cytidines into uracils, thereby creating DNA mismatches that are ultimately processed into mutations or DNA breaks (Chaudhuri et al., 2003; Dickerson et al., 2003; Petersen-Mahrt et al., 2002). While AID primarily acts on Ig variable and switch regions, non-Ig genes including BCL6, CD79A, CD79B, CD83, and PAX5 are occasionally targeted by hypermutation in GC B cells (Shen et al., 1998; Liu et al., 2008). AID off-target activity has been implicated in malignant transformation of GC-derived B cell lymphomas and plasmacytomas (Ramiro et al., 2006; Pasqualucci et al., 2008; Takizawa et al., 2008; Chesi et al., 2008; Robbiani et al., 2008). Additional studies suggest that AID together with RAG1/RAG2 enzymes may also contribute to chromosomal translocations at early stages of B cell development (Tsai et al., 2008).
Under physiological circumstances, AID gene expression is induced in mature germinal center or activated B cells by crosslinking of the B cell receptor, CD40, IL4R, and Toll-like receptors (Crouch et al., 2007; Muramatsu et al., 1999). Downstream activation of NF-κB, along with B cell-specific transcription factors PAX5 and E2A are required for transcriptional activation of AID (Gonda et al., 2003; Sayegh et al., 2003; Dedeoglu et al., 2004; Gourzi et al., 2007; Table S1). We have recently demonstrated that oncogenic BCR-ABL1 kinase activity in acute lymphoblastic leukemia with Philaldelphia chromosome (Ph+ ALL) induces aberrant AID expression (Feldhahn et al., 2007). As in germinal centers, AID in Ph+ ALL cells is enzymatically active and can hypermutate BCL6 and Ig genes (Feldhahn et al., 2007). However, whether AID plays a role in malignant progression of Ph+ ALL or other BCR-ABL1 leukemias remains unclear.
Chronic myeloid leukemia (CML), first identified in 1845 (Virchow, 1845), is -like Ph+ ALL-characterized by the Philadelphia chromosome encoding BCR-ABL1 (Rowley, 1973). CML develops from a hematopoietic stem cell and consequently displays multilineage differentiation potential (Calabretta and Perrotti, 2004). If not efficiently treated, CML follows a triphasic clinical course with an initial indolent chronic phase (CML-CP; 5–15 years) followed by an intermediate accelerated phase and eventually a blast crisis of myeloid (CML-MBC; ~60% incidence), B lymphoid (CML-LBC; 30% incidence) or biphenotypic myeloid/lymphoid (~10% incidence) lineage (Calabretta and Perrotti, 2004). Whereas CML-CP can be effectively treated with Imatinib for many years (Druker et al., 2006), CML-LBC is invariably multidrug-resistant and fatal within weeks or months (Druker et al., 2001). The molecular basis of chronic phase to blast crisis transformation is still largely unknown. However, the majority of patients in blast crisis acquire secondary mutations or deletions at ARF, MYC, RB1, AML1, TP53, and RAS genes (Melo and Barnes, 2007), which are believed to accelerate disease progression. In addition, blast crisis progression correlates with acquisition mutations in the BCR-ABL1 kinase, which confer resistance to Imatinib: The frequency of such mutations is at least six times higher in blast crisis CML than in chronic phase CML (Soverini et al., 2006; Table S1). In a number of patients, BCR-ABL1 kinase mutations can be found even prior to treatment with Imatinib (Pfeifer et al., 2007). Despite their clinical significance, the mechanisms responsible for Imatinib resistance and genetic instability in CML remain unresolved. We here investigate the hypothesis that aberrant AID expression contributes to widespread genomic instability, Imatinib resistance, and B lymphoid blast crisis progression of CML.
To determine whether AID is implicated in CML progression into fatal blast crisis, we examined AID expression in patients with co-existing chronic phase CML (CML-CP) and B lymphoid blast crisis CML (CML-LBC). Leukemia cells of either phenotype were clearly distinguished in bone marrow samples from these patients based on myeloid (CP; CD13+CD19-CD34+) and B lymphoid (LBC; CD13-CD19+CD34+) markers. Real-time RT-PCR showed significant expression of AID mRNA in CML-LBC blasts, but little or no expression in CML-CP cells (6–10 fold difference, P = 0.013, Figure 1A). By comparison, AID expression in primary LBC blasts was 5–10 fold lower relative to germinal center (GC) B cells isolated from human tonsils (Figure 1A). Western blot analysis also showed AID protein present in LBC but absent in CML-CP cells (Figure 1B). Significantly, AID was in all cases coexpressed with the B cell lineage-inducer PAX5 (Nutt et al., 1999), which has been shown to promote AID gene transcription (Gonda et al., 2003; Sayegh et al., 2003; Figure 1C). To measure AID protein expression on a per cell basis, we performed intracellular staining with an AID-specific antibody. Compared to GC-derived diffuse-large B cell lymphoma cells (DLBCL) that express high levels of AID (Lenz et al., 2007), AID protein was undetectable in bone marrow samples from patients with myeloid CML-CP (Figure 1D). In contrast, patients with early CML-LBC harbored cells expressing AID at varying levels (3–18%) along with the B cell antigen CD19 (Figure 1D). Surface expression of the hematopoietic progenitor cell antigen CD34 on these cells excluded the possibility of sample contamination with mature, AID+ activated B cells. We conclude that progression of CML-CP into CML-LBC is characterized by aberrant expression of PAX5 and its transcriptional target AID.
We next examined BCR-ABL1-mediated transformation of bone marrow cells from mice carrying an AID-GFP reporter transgene (Crouch et al., 2007). Isolated bone marrow cells from AID-GFP animals were transduced with a retrovirus expressing human BCR-ABL1 (Pear et al., 1998), or a retroviral empty vector control (MSCV). Following transduction, cells were cultured on bone marrow stroma in the presence of either myeloid (IL3, IL6 and SCF) or B lymphoid (IL7) growth factors following classical models of myeloid CML and B lymphoid LBC/Ph+ ALL respectively (Li et al., 1999). Under myeloid culture conditions, neither normal nor BCR-ABL1-transformed myeloid progenitor cells displayed AID-GFP reporter expression (Figure 1E). In contrast, ~5% of the BCR-ABL1-transformed B lymphoid progenitors exhibited strong AID-GFP (Figure 1E). In agreement with previous findings (Gourzi et al., 2006), retroviral infection with empty vector control also induced AID-GFP fluorescence in a small number of B lymphoid progenitor cells (0.7%, Figure 1E). Of note, AID expression in BCR-ABL1 AID-GFP+ leukemia cells was commensurate to splenic B cells activated with LPS and IL4, and about 240-fold higher than in AID-GFP-counterparts (Figure 1F).
To further characterize AID+ B lymphoid leukemias, we analyzed AID-GFP+ and AID-GFP-cells among BCR-ABL1-transformed B lymphoid leukemia and LPS/IL4-activated splenic B cells with Affymetrix 430 GeneChips (Figure 2). A number of genes typically coexpressed with AID in activated B cells were also upregulated in BCR-ABL1 leukemias (Mfap5, Iigp1, Rcan2, Ccl25, Pim2, Tlr7, Ilr2a, Figure 2A). Conversely, a subset of DNA repair genes were strongly upregulated with AID in activated splenic B cells but absent in BCR-ABL1 leukemia cells (Figure 2B). Among these, Brca1, FancD2 and Rad51 (Longerich et al., 2008), Atm and Cdkn2a (Ramiro et al., 2006), and Chek1 (Gourzi et al., 2006) have been previously implicated either in the repair of AID-induced DNA lesions or negative regulation of AID-induced chromosomal translocation events. Thus, BCR-ABL1-transformed B lymphoid leukemia cells express AID in the absence of the DNA repair mechanisms that typically safeguard genome integrity during normal B cell activation in GCs.
In addition to DNA repair molecules, the micro RNA miR-155 acts as a tumor suppressor by protecting B cells from AID-induced chromosomal translocations (Dorsett et al., 2008). This mechanism relies on the ability of mir-155 to negatively regulate AID protein levels and to curb excessive AID activity during B cell activation in GCs (Teng et al., 2008). Deep sequencing analysis of human GC B cells and mouse activated B cells confirmed that expression of AID in these cells is matched by high levels of miR-155 expression (Figure 2C, D). In contrast, human B lymphoid CML-LBC and mouse B lymphoid BCR-ABL1 leukemia cells expressed high levels of AID in the absence of miR-155 (Figure 2C, D). Hence, while AID is readily expressed in B lymphoid CML-LBC, some of the mechanisms that normally constrain AID-mediated genetic instability are not concurrently induced.
To determine whether AID is functional in human CML-LBC, we next assessed somatic hypermutation of immunoglobulin heavy chain (IGHM), BCL6, and MYC genes, which are targeted by AID in germinal center B cells(Shen et al., 1998; Liu et al., 2008; for a schematic of regions selected for sequence analysis, see Figure S1A–C). In contrast to AID-negative CML-CP cells, AID+ CML-LBC cells carried mutations in IGHM, BCL6, and MYC at frequencies comparable to those previously reported for GC B cell-derived lymphoma (Pasqualucci et al., 2001; Figure 3A). To determine whether AID promotes overall genetic instability in BCR-ABL1 tumors, we measured gene copy number alterations (i.e. deletions or amplifications) in 23 primary cases of Ph+ ALL using a 250K NspI SNP array (Figure 3B). Based on AID mRNA levels, Ph+ ALL samples were classified as either AIDhigh (16 cases) or AIDlow (7 cases). SNP analysis revealed a higher frequency of gene copy number alterations in the AIDhigh as compared to the AIDlow group (median 14 [range 6-50] vs. median 5 [range 2–8]; P = 0.02; Figure 3B). Notably, deletion frequencies at the tumor suppressor genes ARF (CDKN2A) and INK4B (CDKN2B) at 9p21 were considerably higher in the presence of AID (P = 0.04; Figure 3B).
Deletion of ARF has previously been implicated in the progression of myeloid CML-CP into B lymphoid CML-LBC (Calabretta and Perrotti; 2004). To determine whether AID directly targets ARF in LBC cells, we assayed this locus by ligation-mediated PCR for the presence of ssDNA breaks, which are intermediates of AID-mediated hypermutation(Unniraman and Schatz, 2007). We found evidence of ssDNA breaks at the ARF locus of sorted AID+ LBC cells, whereas these lesions were absent in AID-CML-CP cells isolated from the same patients (Figure 3C), indicating ARF is a potential AID target in LBC cells. In support of this view, the CDKN2A (ARF) gene promoter carries five E box motifs (caggtg). E box motifs are tightly associated with AID hypermutation activity at Ig and non-Ig genes alike (Michael et al., 2003). In addition, analysis of a database of published ARF mutations in B lymphoid Ph+ ALL and CML-LBC revealed a significant preference for mutations at RGYW or WRCY DNA motifs (13 of 60 mutations (21.7%), P<0.05), which are hotspots for AID-induced hypermutation (Dorner et al., 1997). No such preference was observed in non-lymphoid tumors (24 of 461 mutations (5.2%), P>0.22; Figure S2). These observations suggest that AID promotes genetic instability in CML-LBC via widespread somatic hypermutation. To further explore this idea, we transformed bone marrow B cell precursors from AID−/− and AID+/+ mice with BCR-ABL1-expressing retroviruses. After 9 weeks in cell culture, we assessed genomic deletions and amplifications by comparative genomic hybridization (CGH). The frequency of copy number alterations was significantly higher in AID+/+ B lymphoid leukemia cells (52 ± 12) as compared to their AID−/− counterparts (20.6 ± 2.1; P=0.02; Figure 3D). Of note, 22 genes were consistently deleted in the presence but not in the absence of AID (Figure 3E). Among these, we found examples of genes implicated in the DNA repair and/or DNA damage response: Deletions of the Msh3, Ddi2 and Ctnnd2 genes were recurrently found in AID+/+ but not AID−/− B lymphoid BCR-ABL1 leukemias (Figure 3E), suggesting that AID promotes genetic instability in BCR-ABL1 lymphoid leukemias in part via hypermutation of DNA repair and DNA damage response genes. Acquisition of BCR-ABL1 kinase domain mutations represents a critical event in the progression of CML-CP into fatal CML-LBC, as they often confer resistance to Imatinib treatment(Shah et al., 2002). To investigate a potential role of AID in this process, we amplified and sequenced the BCR-ABL1 kinase domain from single AID-negative myeloid and AID+ B lymphoid CML cells isolated from four patients with early CML-LBC under Imatinib-treatment. Single-cell PCR analysis and direct sequencing revealed the presence of L248V and T315I mutations in a significant fraction of B lymphoid but not myeloid CML cells in three of the patients (Figure 4A). The fourth patient showed positive PCR products in only a few sorted B lymphoid CML cells. However, the E255K mutation was repeatedly amplified from bulk CML-LBC cells whereas myeloid cells from the same patient were unmutated (not shown). Since these findings show a straight correlation between AID expression and the acquisition of clinically relevant BCR-ABL1 mutations in CML-LBC, we investigated whether enforced AID expression can induce BCR-ABL1 mutations in CML. To this end, we transduced eight AID-negative myeloid CML cell lines with a retroviral vector encoding AID/GFP or GFP alone (Figure S3).
Transduced cells were cultured in the presence or absence of Imatinib for three weeks and the relative enrichment or depletion of GFP+ cells was monitored by flow cytometry. While GFP+/GFP-ratios were unchanged in untreated cultures, we observed a preferential expansion of AID-GFP transduced CML cells in the presence of Imatinib (Figure 4B). Importantly, single-cell PCR analysis consistently showed the presence of clinically relevant BCR-ABL1 kinase mutations in Imatinib-treated AID/GFP+ but not GFP+ CML cells (Table 1). These results were confirmed in a complementary assay based on enzymatic digestion of BCR-ABL1 kinase amplification products (Figure S4). We conclude that enforced AID expression can promote mutagenesis of the BCR-ABL1 oncogene in CML cells. To examine whether this activity leads to Imatinib-resistance in vivo, we then labeled GFP+ and AID/GFP+ transduced CML cells with lentiviral firefly luciferase and injected them intrafemorally into sublethally irradiated NOD/SCID mice. Leukemia engraftment was first observed 12 days post-injection (Figure 4C), as determined by bioluminescence imaging. When focal expansion of leukemic growth became evident (day 20), treatment with Imatinib (100 mg/kg twice daily) was started. By day 35, all 14 mice injected with AID/GFP-transduced CML cells had developed progressive disease as opposed to 5 of 12 GFP control mice (P <0.02, Figure 4C). Correspondingly, all mice injected with AID/GFP+ CML cells were dead at day 54, whereas about half of the mice injected with GFP+ CML cells were still alive 162 days post-injection (Figure 4D).
Our transduction studies clearly show that while not absolutely required, AID can accelerate the acquisition of BCR-ABL1 kinase domain mutations in human CML. To test AID-dependence of these mutations in a genetic experiment, we transduced bone marrow B cell precursors from AID+/+ and AID-mice with BCR-ABL1 to generate AID+/+ and AID-B lymphoid BCR-ABL1 leukemias. AID+/+ and AID-B cell precursors transformed by BCR-ABL1 at a similar efficiency, showed similar leukemic growth kinetics, viability, and colony formation potential in a methyl-cellulose semi-solid culture system (Figure S5). Six weeks following transformation, AID+/+ and AID-leukemia cells were treated with increasing concentrations of Imatinib (from 0.1 up to 1.75 μmol/l) over a period of two weeks. Under these conditions, we observed a reproducible survival advantage of AID+/+ relative to AID-leukemia cells (not shown). Importantly, sequence analysis of the BCR-ABL1 kinase domain revealed multiple AID+/+ clones carrying BCR-ABL1 kinase mutations commonly observed in Imatinib-resistant LBC (Shah et al., 2002; Table S2; Figure 4E). Conversely, the BCR-ABL1 mutation frequency in AID−/− leukemias was within the range of Pfu DNA polymerase error rate (P = 0.09; Figure 4E). A significant number of mutations in AID+/+ cells were C→T (e.g. T315I) or G→A (e.g. E255K) transitions, suggesting that AID could directly target the BCR-ABL1 fusion gene.
The transcription factor PAX5 regulates both AID-gene expression (Gonda et al., 2003; Sayegh et al., 2003; Table S1) and B cell lineage commitment of hematopoietic progenitor cells (Nutt et al., 1999). To investigate whether ectopic expression of PAX5 in myeloid CML cells was sufficient to induce AID transcription and B cell lineage conversion, we transduced PAX5-negative CML cells with a retroviral vector expressing PAX5/GFP or GFP alone. Enforced expression of PAX5 resulted in significant upregulation of known B cell-specific PAX5-target genes including AID, SLP65 (BLNK), and CD79A (Igα; Figure 5A). This was accompanied by surface expression of CD19 in a small subset of PAX5/GFP+ but not in GFP+ CML cells (Figure 5B). Upon prolonged treatment with Imatinib, CD19+ CML subclones were positively selected to more than 16% after 6 weeks (Figure 5B). Furthermore, immunoglobulin spectratyping analysis of PAX5/GFP-transduced CML cells showed evidence of PAX5-induced de novo immunoglobulin VH-DJH rearrangement (Figure 5C), which is consistent with the role of PAX5 at promoting VH-DJH recombination (Zhang et al., 2006). Sequence analysis of immunoglobulin gene rearrangements in PAX5/GFP-transduced CML cells revealed a striking preference for the most proximal located VH gene segments: among 37 informative clones, only VH6-1 and VH1-2 gene segments (#1 and #2 most proximal located among 123 VH segments) were found rearranged. PAX5-induced V(D)J recombination was also aberrant: instead of an initial DH-JH joint rearranging to different VH-segments, a pre-existing VH-DH joint (VH6-1-DH5-5) was rearranged to various JH segments (JH4 and JH6; Figure 5C). We conclude that ectopic expression of PAX5 in myeloid CML cells induces partial B cell lineage conversion and AID expression.
To determine whether PAX5 can promote Imatinib-resistance in CML, PAX5/GFP+ and GFP+ human leukemia cells were cultured in the presence of increasing concentrations of Imatinib (0.1 μmol/L to 1.75 μmol/L). In analogy to results obtained with AID/GFP(Figure 4B), we observed a time-dependent outgrowth of PAX5/GFP+ CML cells in the presence of Imatinib (Figure 5D). Sequence analysis confirmed the presence of mutations within the BCR-ABL1 kinase domain of PAX5/GFP but not in GFP-transduced CML cells (Table 2). We conclude that enforced PAX5 expression in myeloid CML cells can lead to partial B cell lineage conversion, AID expression, and Imatinib-resistance via BCR-ABL1 mutation.
AID-mediated hypermutation preferentially targets RGYW and WRCY hotspot motifs (Dorner et al., 1997) and displays a biased towards C→T and G→A transitions over transversion mutations (Di Noia and Neuberger, 2002; Faili et al., 2002). To explore whether BCR-ABL1 is a direct target of AID, we assembled and analyzed a database of 700 published BCR-ABL1 kinase domain mutations (572 from CML-CP and 128 from CML-LBC/Ph+ ALL; Figure S6; Table S2). In support of direct AID-mediated hypermutation of BCR-ABL1, we found the frequency of C→T and G→A transitions to be notably higher in B lymphoid CML-LBC/Ph+ ALL (66.4%) relative to CML-CP (19.9%; P<0.01; Figure 6B; Table S2). This difference was exemplified by the 763G→A transition leading to E255K, and the 764A→T transversion leading to E255V. Both mutations target the same codon and are associated with a similar degree of Imatinib-resistance (IC50 [Imatinib] at 12.1 μmol/l and 17 μmol/l, respectively). In contrast to E255V, the E255K G→A transition occurs 5 times more frequently in CML-LBC/Ph+ ALL than in CML-CP. The E255V A→T transversion however, was found in 22 cases of CML-CP (3.8%), but not in a single case of CML-LBC/Ph+ ALL (Table S2). We next investigated whether BCR-ABL1 kinase mutations occurred preferentially at RGYW/WRCY motifs based on a previous cytosine deamination quantitative assay, which determined the ability of 68 cytosine-based DNA motifs to serve as a substrate for AID (Yu et al., 2004). We found a significant correlation between potential AID activity (percentage in vitro deamination) and BCR-ABL1 mutation frequency in CML-LBC/Ph+ ALL (r=0.36; p=0.01; Figure 6C, D; Figure S6). In contrast, no such correlation was observed for CML-CP (r=−0.14; p=0.36; Figure 6E). Importantly, statistical analyses showed no significant correlation between IC50 values and AID targeting (r=0.04; p=0.84), a feature that argues against the possibility that our results are biased by selective advantage of individual mutations (see Figure 6F and S6 for detailed explanation).
We have shown here that the transition from chronic phase CML into B lymphoid blast crisis is accompanied by expression of PAX5 and its transcriptional target, the B cell-specific mutator AID. The presence of hypermutation at MYC, ARF, BCL6, and the IGHM loci clearly demonstrates the enzymatic activity of AID in CML-LBC. In like manner, a subset of BCR-ABL1-induced mouse leukemia cells express very high levels of AID and carry deletions and amplifications at an increased frequency relative to BCR-ABL1-induced leukemia cells from AID−/− mice. The first hint to AID tumorigenic activity was given by studies of transgenic mice expressing AID constitutively and ubiquitously under the chicken β-actin promoter (Okazaki et al., 2003). These animals rapidly succumbed to T cell lymphomas, lung microadenomas, and to a lesser extent sarcomas, hepatocellular carcinomas and melanomas. Surprisingly, B cell ontogeny was not directly affected by AID overexpression. Most importantly, B cell malignancies were not detected in these mice, suggesting that B lymphocytes might have evolved B cell-specific mechanisms to suppress AID tumorigenic activity (Casellas et al., 2009). Error-free base excision and mismatch repair pathways for instance, correct the vast majority of AID-mediated mutations outside the immunoglobulin loci in germinal center B cells (Liu et al., 2008). Likewise, DNA repair proteins, cell cycle regulators, and microRNA mir155 prevent AID-induced DNA double-strand breaks from developing into chromosomal translocations (Dorsett et al., 2008). In stark contrast to activated or GC B cells, B lymphoid AID+ CML-LBC cells fail to upregulate the full repertoire of protective mechanisms that normally accompany AID activity. This feature, which is likely to be a recurrent theme in AID+ malignancies, would exacerbate AID-mediated hypermutation and/or deletion of tumor suppressor genes. Based on these observations, we propose that aberrant activity of AID and the lack of mechanisms restricting its ability to introduce deletions and chromosomal translocations combine to accelerate the progression from chronic phase into fatal B lymphoid blast crisis in CML patients.
In addition to genomic instability, our data uncovered a correlation between AID expression, the acquisition of BCR-ABL1 kinase domain mutations and Imatinib resistance in CML-LBC. This correlation is underscored by the observation that BCR-ABL1 kinase mutations are at least six times more frequent in B lymphoid CML-LBC than in myeloid CML-CP (Soverini et al., 2006; Table S1). Three sets of experiments argue for a causative role of AID in the acquisition of BCR-ABL1 kinase mutations: i) forced expression of AID induces BCR-ABL1 kinase mutations and Imatinib-resistance in human CML cells; ii) forced expression of PAX5 in CML cells leads to AID gene transcription and the acquisition of BCR-ABL1 kinase mutations and iii) AID+/+ but not AID−/− mouse BCR-ABL1 leukemias acquire BCR-ABL1 mutations and Imatinib-resistance within a short period of time. While BCR-ABL1 mutations do obviously occur in the absence of AID (e.g. as in some cases of CML-CP), our findings demonstrate that AID dramatically accelerates the acquisition of such mutations and, hence, increases their overall frequency (e.g. as in CML-LBC; Table S1).
In this context, BCR-ABL1 mutations could arise either directly by way of direct AID-targeted hypermutation, or indirectly as a result of AID-mediated genetic instability (e.g. by targeting of genes that are involved in DNA repair). Below we explore these arguments in detail.
In germinal center B cells, somatic hypermutation occurs mostly within a 2 kb window downstream of a gene’s transcription start site (TSS), and loses its activity exponentially with increasing distance from this window (Rada and Milstein, 2001). Depending on the precise location of the breakpoints within the BCR and ABL1 genes, the BCR-ABL1 kinase domain is situated approximately 60 to 75 kb downstream of the BCR TSS (Figure S1E), a feature that argues against the direct targeting hypothesis. On the other hand, hypermutation of immunoglobulin switch domains is known to occur up to 10 kb downstream of sterile promoters during isotype switching. Hence, if recruited to an unknown transcription start site downstream of the BCR promoter, AID could in principle exhibit enough residual activity to target the BCR-ABL1 kinase domain. Since many BCR-ABL1 kinase mutations confer drug-resistance, such lesions will be eventually selected from a heterogeneous leukemia cell population even if introduced at extremely low frequencies. Attempts to uncover hypermutation near the BCR promoter in CML-LBC cells have been inconclusive so far (our unpublished observations). A negative result, however, does not necessarily rule out direct targeting of BCL-ABL1 by AID, as hypermutation of a large number of genes in germinal center B cells is only evident in the absence of error-free repair proteins Msh2 and UNG1 (Liu and Schatz, 2008). Alternatively, AID hypermutation could be recruited by the QRFP, LOC100131443 or FIBCD1 genes, which are actively transcribed at 5.4 kb, 10.4 kb and 51.1 kb downstream of and in opposite orientation vis-a-vis BCR-ABL1 (Figure S1E).
Of note, the BCR-ABL1 fusion gene is located at chromosome 22q11 in close proximity to the human immunoglobulin λ locus (Figure S1E), which strongly recruits AID. The ability of the immunoglobulin λ locus to recruit AID was further underscored by the recent identification of a cis-acting Diversification Activator (DIVAC) element, which induces AID-dependent hypermutation of both immunoglobulin and non-immunoglobulin genes in DT40 chicken B cells (Kothapalli et al., 2008; Blagodatski et al., 2009; Figure S1E). DIVAC encompasses the immunoglobulin light chain (IgL) enhancer in chicken, a region that is highly homologous to the human immunoglobulin λ locus (IGLV), which is located in close proximity to the BCR-ABL1 fusion gene (Figure S1E). It is conceivable that enhanced transcriptional activity following the chromosomal break event at 22q11 in the regional context of both the immunoglobulin λ locus and DIVAC-like elements may lead to recruitment of AID and aberrant hypermutation of the BCR-ABL1 fusion gene.
Although based on correlative observations, the pattern of BCR-ABL1 kinase mutations in CML-LBC/Ph+ ALL indeed suggests direct targeting by AID at a low frequency (Figure 6). Our data however do not rule out that mechanisms downstream of or independent from AID are also at play. Genomic instability resulting from AID-mediated deletion of DNA repair genes would be an example of the former, the acquisition of random mutations as byproducts of cellular oxidative stress or DNA replication errors would exemplify the latter. For instance, a recent mathematical model for drug-resistance in CML proposes that DNA replication errors lead to the acquisition of BCR-ABL1 kinase mutations at a rate of 4 × 10−7 per cell division (Michor et al., 2005). This model explains the high frequency of BCR-ABL1 kinase mutations in CML blast crisis on the basis of the particularly high proliferation rate of these cells.
CML develops from a hematopoietic stem cell and consequently displays multilineage differentiation potential. During the chronic phase of the disease, the vast majority (>95%) of tumor cells differentiate into committed myeloid progenitors, while a small fraction (<5%) develops into the B lymphoid lineage likely as a result of PAX5 upregulation (Takahashi et al., 1998; Verstegen et al., 1999). In this context, we have shown that enforced expression of PAX5 promotes B cell lineage conversion in CML cells and induces aberrant expression of AID. We propose that this is a key step in the progression from CML-CP to fatal CML-LBC.
The development of Imatinib resistance in CML-MBC, which are PAX5-AID-(our unpublished observations), argues against a role of AID in the acquisition of BCR-ABL1 mutations in myeloid blast crisis. However, CML typically exhibits lineage infidelity and in ~10% of cases, B/myeloid biphenotypic leukemia cells give rise to CML blast crisis. By oscillating between myeloid and B lymphoid phenotypes (Takahashi et al., 1998; Verstegen et al., 1999), Imatinib-resistant MBC could in principle originate from mutated B lymphoid CML cells that have subsequently lost their B cell identity. There are several precedents supporting this view: MYC-driven hematopoietic progenitor cell leukemias are known to alternate between B lymphoid and myeloid lineages through spontaneous silencing and reactivation of PAX5-dependent B cell lineage-induction (Yu et al., 2003). Furthermore, Hodgkin’s disease represents a prominent example of a human B cell-derived malignancy that has lost its B cell identity through aberrant expression of ID2, an inhibitor of PAX5/E2A (Mathas et al., 2006).
Interestingly, we have found that in nearly all cases of CML-MBC analyzed, IGHM VH-DJH and the Ig κ deleting element (KDE) are rearranged (Figure S7). Rearrangement of these loci is considered as a permanent imprint of B cell identity that would be retained even after subsequent myeloid lineage conversion. Importantly, VH-DJH genes in CML-MBC show evidence of somatic hypermutation (Figure S7), whereas the IGHM and KDE loci are unmutated and in germline configuration in CML-CP. These findings support a B lymphoid origin for at least some cases of CML-MBC, and raise the intriguing possibility that AID activity may promote blast crisis transformation also in the case of CML-MBC.
Normal naïve B cells (CD19+ CD27-IgD+) were sorted from peripheral blood of 12 healthy donors by flow cytometry using a FACSVantage SE cell sorter (BD Biosciences). Human germinal center B cells (CD77highCD38+IgD−) were isolated from tonsillar resectates (TKH) and sorted by FACS. Bone marrow samples from patients with CML-CP and early B lymphoid blast crisis CML (Table S3) were provided from the Departments of Hematology and Oncology, Charite University Hospital Benjamin Franklin, Berlin, Germany (WKH) and the GIMEMA Study Group/University of Bologna, Bologna, Italy (GM) in compliance with Institutional Review Board regulations (approval from the Ethik-Kommission of the Charite, Campus Benjamin Franklin and the IRB of the University of Southern California Health Sciences Campus and Committee on Clinical Investigations (CCI) of Childrens Hospital Los Angeles). Informed consent was obtained from all human subjects. Human CML cell lines (EM2, JK1, JURL-MK1, K562, KCL22, KYO, LAMA84, MEG1) and B cell lymphoma cell lines (Karpas-422, MN60, DB) were obtained from DMSZ, Braunschweig, Germany.
Aid−/− Balb/c mice were derived from Aid−/− mice (Muramatsu et al., 2000) by speed congenic backcrossing and kindly provided by Dr. Michel C. Nussenzweig (New York). Aid-GFP reporter-transgenic mice were described in Crouch et al., 2007. Mice studied in the experiments described here carried two copies of the Aid-GFP reporter transgene. A detailed description of our CML xenograft model is provided in the Supplementary Materials section. All experiments involving mice conform to the relevant regulatory standards and were reviewed and approved by Childrens Hospital Los Angeles Institutional Animal Care and Use Committee (IACUC).
Quantitative real-time PCR carried out with the SYBRGreenER mix from Invitrogen (Carlsbad, CA) was performed according to standard PCR conditions. During the PCR amplification, the SYBRGreenER dye in the mix binds to accumulating double-stranded DNA and generates a fluorescence signal proportional to the DNA concentration which can be visualized and measured using a ABI7900HT (Applied Biosystems, Foster City, CA) real-time PCR system. Primers for quantitative RT-PCR are listed in Table S4.
Mouse CD43− naïve B cells were activated ex-vivo in the presence of LPS/IL4 for 72 hours. Human germinal center B cells (CD77highCD38+IgD−) were isolated from tonsil samples by cell sorting. Total RNA was isolated from 106 cells and microRNA was processed for Illumina’s deep sequencing using the manufacturer protocol (Illumina, San Diego, CA). Mature microRNA sequences (tags) were quantified based on 1 million tags aligned to the mouse and human microRNome as defined by mirBase (http://microrna.sanger.ac.uk/sequences/search.shtml).
PAX5/GFP and GFP-transduced CML and primary CML cells were analyzed for presence of IGHM and KDE gene rearrangements (Table 3) and clonality by spectratyping. To this end, RNA was extracted and subjected to cDNA synthesis. From cDNAs, VH-DJH gene rearrangements were amplified using PCR primers specific for VH gene segments together with a primer specific for the Cμ constant region (Table S4). Using a FAM-labeled Cμ constant region gene -specific primer in a run-off reaction, PCR products were labeled and subsequently analyzed on an ABI3730 DNA analyzer by fragment length analysis. A normal polyclonal B cell repertoire exhibits a bell-shaped curve of individual size-peaks according to a Gaussian-type distribution of length diversity within the VHDJH junction.
For mutation analysis of ARF, BCL6 and MYC genes, genomic fragments were amplified and sequenced using Pyrococcus furiosus DNA polymerase (New England Biolabs, Ipswich, MA). For each PCR product, both DNA strands were sequenced and mutations were only counted if they were found both in the forward and reverse sequence. PCR primers used for amplification of ARF, BCL6 and MYC fragments are listed in Table S4. The ABL1 kinase portion of BCR-ABL1 transcripts was amplified in two rounds of PCR: To prevent co-amplification of normal ABL1 transcripts, the first round of PCR used BCR- (exon 13) and ABL1- (exon 9) specific primers in 10 cycles of amplification. The second round of amplification (35 cycles) focused on the ABL1 kinase domain (exons 4-6). The rate of Pfu DNA polymerase errors was calculated at 4 × 10−5/PCR cycle using genomic DNA from MACS-sorted CD3+ human T cells (immunomagnetic beads from Miltenyi Biotech). A detailed description of the analysis of DNA single-strand breaks by ligation-mediated PCR is provided in the Supplementary Materials section.
Antibodies against human CD19, CD13, CD34, CD38, IgD, VpreB, CD79B (Igβ), IL7α and μ-chain as well as antibodies against mouse CD19, B220, IL7rα, Gr1, c -kit, Sca-1, AA4.1 as well as respective isotype controls were purchased from BD Biosciences, Heidelberg, Germany. For analysis of AID expression by flow cytometry, antibodies against AID (rat anti-human AID IgG2b κ, clone EK2 5G9, Cell Signaling Technology, Beverly, MA) were used together with secondary antibodies (goat anti-rat IgG Cy2, Jackson ImmunoResearch, West Grove, PA). Prior to staining, cells were fixed with 0.4% para-formaldehyde and incubated for 10 minutes in 90% methanol on ice. For the detection of AID by Western blot, an antibody against human AID (L7E7; Cell Signaling Technology), was used together with the WesternBreeze immunodetection system (Invitrogen, Karlsruhe, Germany). Detection of EIF4E was used as a loading control (Santa Cruz Biotechnology, Santa Cruz, CA).
AID/GFP- or GFP-transduced CML cells were FACS-sorted into 0.5 ml reaction tubes containing 5μl of 2X ribonuclease inhibitor mix (2μl RNasin [40u/μl] + 18μl 0.15 mol/l NaCl/10 mmol/l Tris-HCl [pH 8.0]/5mM DTT). The cells were then quickly frozen on dry ice and stored at −80°C. After one freezing and thawing step, which leads to cell lysis, the 10 μl cell lysates were added to 50μl RT-PCR using the Access RT-PCR System (Promega, Madison, WI). RT-PCR primers for BCR-ABL1, ABL1 or COX6B (Table S4), were used (50 pmoles each per reaction). To prevent co-amplification of normal ABL1 transcripts, the first round of PCR used BCR (exon 13)- and ABL1 (exon 9)-specific primers for specific amplification of BCR-ABL1 fusion transcripts. The second round of amplification focused on the [BCR-] ABL1 kinase domain (exons 4–6). The rate of Taq DNA polymerase errors was calculated at 2 × 10−4/ PCR cycle and PCR products were directly sequenced.
Bone marrow progenitor cells from AID-GFP reporter transgenic mice were transformed with a retroviral BCR-ABL1 vector under B lymphoid growth conditions (IL7). As a control, splenocytes were isolated from AID-GFP reporter transgenic mice and incubated in the presence of 25 μg/ml LPS and 10 ng/ml IL4. We sorted AID-GFP+ and AID-GFP- cells from BCR-ABL1-transformed leukemia cells and LPS/IL4-stimulated splenocytes and subjected the cells to RNA isolation and microarray analysis using the 430 GeneChip platform (Affymetrix, Santa Clara, CA).
Genomic DNA was extracted from AIDhigh and AIDlow Ph+ ALL samples and subjected to 250K NspI SNP array analysis as described in the Supplementary Materials section. For comparative genomic hybridization, bone marrow cells from AID−/− and AID+/+ mice were transduced with BCR-ABL1. After 9 weeks in cell culture, genomic DNA was isolated and amplifications and deletions were detected using the Mouse Genome CGH Microarray Kit 244A (Agilent Technologies, Inc., Santa Clara) as described in the Supplemental Materials section.
Microarray gene expression data, SNP Chip data and CGH data are available from GEO. Microarray gene expression data on BCR-ABL1 leukemia cells from Aid-GFP transgenic bone marrow are available under GEO accession number GSE13611. SNP array data on 16 AIDhigh and 7 AIDlow cases of Ph+ ALL are available under GEO accession number GSE13612. CGH genome tiling array data from AID+/+ and AID−/− BCR-ABL1-transformed leukemia cells are available under GEO accession number GSE15093.
Treatment of patients with CML in chronic phase with the BCR-ABL1 kinase inhibitor Imatinib leads to a five-year overall survival of ~95%. Upon blast crisis progression, however, median overall survival decreases to less than seven months and CML blasts are frequently Imatinib-resistant due to BCR-ABL1 mutations. Here we show that PAX5-mediated B cell lineage conversion of CML cells results in transcription of AID, a B cell-specific enzyme that hypermutates immunoglobulin and other genes in germinal center B cells. In CML cells, AID promotes widespread hypermutation, B lymphoid blast crisis progression, BCR-ABL1 mutation and the acquisition of Imatinib-resistance. Our studies shed light on the origin of CML-LBC and identify AID as a potential pharmacological target to suppress fatal blast crisis transformation.
We would like to thank Michel C. Nussenzweig (New York) and Tasuku Honjo (Kyoto) for sharing AID−/− Balb/c mice with us. We are indebted to Tanja A. Gruber for help with CGH analysis, Bonaventure Ndikung Soh for KDE analysis; Kirsten Schneider and Jessica Langer for help with sequencing. This work is supported by grants from the NIH/NCI through R01CA137060 (M.M.), R01CA139032 (M.M.), the Leukemia and Lymphoma Society (to M.M.), the V Foundation for Cancer Research (M.M.), the Deutsche Forschungsgemeinschaft (German Science Foundation) through MU1616/5-1 and the Kenneth T. and Eileen L. Norris Foundation (J.G., M.M.).
The authors have no conflicting financial interests.