AID overexpression leads to aggressive murine CLL and nonimmunoglobulin mutations that mirror human neoplasms

Chronic lymphocytic leukemia (CLL) results from the accumulation of mature, long-lived CD5⁺ B lymphocytes in peripheral blood (PB), bone marrow, and secondary lymphoid organs^1,2  due to an imbalance between cell proliferation and death.³  Furthermore, disease progression correlates with the relative proportions of proliferative and quiescent subpopulations within leukemic clones.^4,5  Activation of CLL cells mainly occurs in secondary lymphoid organs,^6,7  where they are stimulated via B-cell receptors,⁸  CD40, and/or Toll-like receptors.^9,10  In B lymphocytes, such receptor/ligand interactions lead to expression of activation-induced cytidine deaminase (AID), an enzyme required for somatic hypermutation and class switch recombination.¹¹  AID mutates genomic DNA by deaminating cytosine to uracil, which can lead to point mutations and DNA double-stranded breaks.^12,13  Although AID preferentially acts on immunoglobulin genes, it can also target other genes giving rise to mutations and/or chromosome translocations that can be oncogenic.¹⁴ 

We have shown that leukemic clones in patients with worse clinical outcomes overexpress AID, and that AID expression is modulated by microenvironmental signals.^15-17  Despite these and other studies¹⁸  suggesting a link between AID activity and CLL progression, direct experimental evidence indicating that AID is responsible for off-target mutations driving disease progression is not available.

Here, we tested the hypothesis that mice overexpressing AID in leukemic cells would recapitulate the leukemogenic process observed in progressive and immunoglobulin heavy chain (IGHV)-unmutated CLL (U-CLL) clones expressing AID. To address this, we crossed Eμ-TCL1 mice (a validated murine model of CLL^19,20 ) with 2 different strains overexpressing AID, either only in B cells¹²  or ubiquitously.²¹ 

Our results show that these double-transgenic AID (DT-AID) mice develop more aggressive leukemias with shorter survivals than Eμ-TCL-1, linking constitutive AID expression in the leukemic clone with enhanced cell proliferation. Whole-exome sequencing (WES) of leukemic cells from DT-AID and their proliferating fractions reveal novel AID-induced mutations in oncogenes and tumor-suppressor genes as well as in genes involved in chromatin structure and the Wnt-signaling pathway. Furthermore, we found that some genes that are mutated in human CLL and diffuse large B-cell lymphoma (DLBCL) are also mutated in DT-AID mice, and that some of these mutations occur at the exact same positions, resulting in equivalent amino acid substitutions as in the human disease. Hence, our 2 DT-AID strains provide complementary models of aggressive CLL-like disease and strongly indicate that AID causes a distinct set of off-target mutations in genes that are likely relevant to CLL and human B-cell neoplasms in general because of selection for their oncogenic effects.

PB mononuclear cells were collected in a prospective manner from patients with CLL defined by International Workshop on Chronic Lymphocytic Leukaemia (IWCLL) criteria. Progressive CLL cases were identified as described in Prieto et al.²²  Supplemental Table 1 (available on the Blood Web site) lists the clinical and molecular characteristic of these patients.

Eu-TCL1–transgenic mice¹⁹  were crossed with either AID-transgenic mice under the control of the actin (act-AID^Tg/wt mice) promoter²¹  or AID-transgenic mice under control of the immunoglobulin κ (Igκ; Igκ-AID^Tg/Tg mice) promoter¹²  in C57BL/6 × (C57BL/6 × C3H) or pure C57BL/6 final genetic backgrounds. Genotyping was performed by polymerasce chain reaction from tail tissue using specific primers for TCL1 and AID genes.^19,21 

Total RNA was isolated from purified leukemic cells and retrotranscribed as described.²³  Gene analysis and quantification of AID expression were performed as described,^21,23  and primer sets were depicted in supplemental Materials and methods or in Palacios et al.²³  Immunohistochemistry was performed as described²⁴  and detailed in supplemental Materials and methods.

Genomic DNA was purified from isolated leukemic Ki67⁺ and Ki67⁻ cells (tumor) and tail tips (germline) using the Quick-DNA Kit (Zymo Research). A similar procedure was used for human PB mononuclear cells where the leukemic clone was isolated and the T-cell fraction was used as a germline control. Reads were aligned as described^25,26  and variant calling was performed using VarScan with a minimal variant frequency of 0.08, and minimum coverage of 10× (V2.4.3).²⁷  Mutation signature analysis was performed using SomaticSignatures²⁸  and methods are detailed in supplemental Materials and methods.

Data are presented as means plus or minus standard error of the mean. The 1-way analysis of variance (ANOVA) multiple comparisons test, the 2-tailed, unpaired Student t test, and the Fisher exact test were used for statistical analysis (*P < .05, **P < .001, ***P < .001, ****P < .0001). Kaplan-Meier survival curve plots, the Gehan-Breslow-Wilcoxon test, and the log-rank (Mantel-Cox) test were performed. All tests were performed with GraphPad Prism v.6 software. A value of P < .05 was considered significant.

The strategy to generate DT mouse models is depicted in supplemental Figure 1A. DT-AID mice were healthy, without evidence of inborn alterations. AID messenger RNA (mRNA) levels were significantly higher in the DT-AID models compared with TCL1 (supplemental Figure 1B). Cytoplasmic AID protein was increased within B220⁺ cells in the DT-AID models compared with the TCL1 counterpart (Figure 1A). All 3 strains showed loss disruption of lymphoid tissue architecture and weaker B220 signals compared with normal B-cell follicles. Disease kinetics were monitored every 2 months by white blood cell counts, proportion of IgM⁺CD5⁺ cells, and associated signs of leukemia. Mice of 6 and 10 months of age were selected as these time points were illustrative of preleukemic and leukemic stages, respectively.²⁹  In both DT-AID models, circulating leukemia cells appeared earlier and at 10 months were significantly higher than in TCL1 (Figure 1B-C; supplemental Figure 1C). Tumor burden was higher, with significantly larger spleens and more IgM⁺CD5⁺ cells in the DT-AID than in TCL1 (Figure 1C-D), whereas both DT-AID models had significantly shorter survival than TCL1 (Figure 1E; supplemental Figure 1E). As expected based on the constitutive expression of AID, many more IGHV mutations were found in the B-cell expansions identified in the DT-AID mice than in the TCL1 animals (supplemental Figure 2).

Genome-wide DNA mutations induced by AID overexpression were identified (supplemental Excel File 1, sheets 1-6). Since the clinical findings, variant rates and mutation patterns were similar in both DT-AID mouse lines (Figure 2; supplemental Figure 3A; supplemental Excel File 1, sheet 1), they were considered together for all subsequent analyses. WES analysis revealed a mutation rate in protein-coding regions of 0.83 ± 0.88 per Mb in TCL1, similar to the mutation rate reported for human CLL 0.6 ± 0.28.³⁰  As expected, AID overexpression in DT-AID mice led to a higher mutation rate (1.59 ± 0.83 per Mb).

Analysis of the somatic mutation pattern³¹  showed a higher proportion of C>T transitions at the preferred RCY AID hotspots in DT-AID than TCL1 mice (Figure 2A; supplemental Figure 3B). Unbiased assessment of underlying mutagenic processes showed that 3 main signatures accounted for >95% of the variance in the sample set, with a higher contribution of the AID signatures in the DT-AID than the TCL1 mice (Figure 2B). Interestingly, mutations at AID canonical hotspots (c-AID; C>T/G within RCY motifs)³²  dominated (supplemental Figure 3).

We identified mutations that might have resulted from the action of AID using 3 criteria: (1) those occurring at c-AID, (2) those considered to be indirectly induced by AID through noncanonical mismatch repair (nc-AID; A>C/G at WAN motifs),³³  and (3) C>T transitions in CpG (NCG).³⁴  These variants are collectively referred to as “AID mutations” hereafter and were more abundant in DT-AID than TCL1 (Figure 3C; supplemental Excel File 1, sheet 5). However, of these 3, c-AID mutations were the only signature significantly increased in DT-AID mice, either genome-wide (Figure 2D), or specifically in exons, noncoding RNA (ncRNA), and 5′ untranslated region (UTRs) (Figure 2E).

Next, we defined those genes that were recurrently mutated at c-AID hotspots in the DT-AID strains as those exhibiting ≥5 mutations distributed between at least 2 mice, and not present in the TCL1 model (supplemental Excel File 1, sheets 5-6). Thirty-four genes met these criteria and, among these, 18 genes (53%) have been identified as AID nonimmunoglobulin targets in murine B cells^35-38  (Figure 2F). Furthermore, we identified a series of recurrently mutated genes with a c-AID signature that have not been previously described in either AID-³⁸  or TCL1-transgenic models³⁹  (supplemental Data 1). Thus, the DT-AID mice accumulated c-AID mutations in different cancer-specific genes that were directly mutated by AID. These included a striking accumulation of mutations in genes encoding histones (Hist1h1c, Hist1h1d, Hist1h1e, and Hist2h2aa1) and in tumor-associated transcription factors (Gata3, Bhlhe41, Sox4, Klf2, Zfp36l2, Tcf7). Furthermore, we found a higher number of c-AID mutations in genes associated with tumor development (Myc, Ly6e, Dusp2) as well as linked with CLL development and progression^40-42  (Pim-1, Nfatc1, Mcl1, Pten, and Lef-1) (Figure 2F).

AID expression is restricted to a subpopulation of cells that are dividing⁴³  or have recently divided.⁴⁴  To investigate whether the more aggressive disease in the DT-AID models correlated with tumor proliferation and/or survival rates, we evaluated Ki67 and BCL-2 protein levels in circulating and/or splenic IgM⁺CD5⁺ cells. DT-AID mice had significantly higher percentages of IgM⁺CD5⁺Ki67⁺ cells in PB and spleen than TCL1 and C57BL/6 animals (Figure 3A). In contrast, BCL-2 levels in the same cell populations were not significantly different (Figure 3B).

Functional annotation of genes bearing nonsynonymous mutations in c-AID hotspots in the DT-AID strains identified 7 significantly enriched pathways, most of them involved in cell proliferation and tumor growth. Interestingly, among the 3 path-ways with the highest gene counts, we found Wnt signaling (Panther Pathway Accession [PPA]: P00057), which was associated with CLL progression⁴⁵  (Figure 3C-D). Notably, many of the genes involved in these 3 pathways were shared between the different DT-AID animals and previously linked with CLL progression (Mcl1,⁴⁶ Myc,⁴⁷ Pten,⁴⁸ Nfatc1,⁴¹  and Lef-1⁴² ) or with leukemia in general (Sox4⁴⁹  and Prkch⁵⁰ ) (Figures 2F and 3D, red asterisks).

Because AID’s mutagenic activity is operative in dividing cells⁴³  and associates with tumor development,⁵¹  and given the higher percentage of Ki67⁺ leukemic cells observed in DT-AID mice, we searched nonimmunoglobulin genes bearing AID mutations that could affect leukemia cell proliferation. We isolated splenic IgM⁺CD5⁺Ki67⁺ cells and IgM⁺CD5⁺Ki67⁻ cells from selected mice (Figure 3A, arrows) for WES. Analyses of these data show that the c-AID mutation signature was significantly greater in the DT-AID than the TCL1 mice (Figure 4A-B).

Next, we focused our analysis to genes in the Ki67⁺ fraction bearing any c-AID mutation that could affect gene expression or function: that is, in the UTRs, up and downstream 2-kb gene regions (∼2 kb), as well as nonsynonymous mutations (n = 136) (supplemental Excel File 2, sheet D). This analysis showed 41 genes (30%) only mutated in the Ki67⁻ fraction (subgroup I), 42 genes mutated in both fractions, (subgroup II), and 53 (39%) only mutated in the Ki67⁺ fraction (subgroup III; Figure 4C). Newly mutated genes in the Ki67⁺ fraction (subgroup II shared with a subset of subgroup I and subgroup III not shared with subgroup I; n = 95, 70% of genes) were remarkably enriched in c-AID mutations affecting histones and B-cell neoplasm gene drivers (Figure 4C, red genes). Sixteen of 167 gene drivers (9%) found in the Ki67⁺ fraction have been described in human CLL^30,52,54  and/or DLBCL⁵³  (supplemental Excel File 2, sheet E). Thus, forcing AID expression in TCL1 mice led to the accumulation of c-AID mutations in a subset of driver genes found in human B-cell neoplasms. Wnt signaling was the most overrepresented pathway affected by c-AID mutations inside of the Ki67⁺ fraction (Figure 4D; supplemental Figure 3).

Among genes identified in the Ki67⁺ fraction (subgroup II and III; Figure 4C), 11 were chosen for in-depth analysis because they had nonsynonymous c-AID mutations in coding regions and/or in UTRs or up/downstream regions. These genes clustered into 3 groups: oncogenes (Pim-1, Mcl-1, and Myc), tumor-suppressor genes (Pax-5, Dusp2, Gata-3, Klf2, and Zfp36l2), and histones (Hist1h1d, Hist1h1e, and Hist1h2h2aa1). Most of the c-AID mutations fell in functional regions in both DT-AID models (Figure 5).

Among the oncogenes, we found mutations in the kinase catalytic domain of the protein kinase–encoding PIM-1, in the Bcl-2 apoptosis-regulator domain of MCL-1 protein, and within the 5′UTR of Myc (Figure 5A). In tumor suppressors, each transcription factor (PAX-5, GATA-3, KLF-2, and ZFP36L2) had mutations in DNA-binding domains (Figure 5B). In the case of histones, we found an unexpected number (>14) of c-AID mutations on the H1 family proteins, including H1.3 and H1.4 variants encoded by Hist1h1d and Hist1h1e, respectively. Mutations in H1.3, H1.4, and H2A affected key amino acid positions. All of them are placed at the conserved DNA-binding domains or affect the AKP helix motif of domain H15 in H1.4, which is essential for this histone family’s function (Figure 5C, red letters).

To further explore the role of AID expression in human CLL progression, and to validate the usefulness of the DT-AID models, we performed WES on leukemic clones from 12 progressive patients with CLL (supplemental Table 1; supplemental Excel File 3), systematically identifying genes mutated in the patients and in our murine models. Patients were selected based on continuous AID expression (AID mRNA⁺ in PB white blood cells every 3 months until treatment), Binet stage, and short time to first treatment. WES data from the human and murine CLL samples were analyzed for mutational context, focusing on putative AID hotspots (c-AID; nc-AID and NCG). Our initial analysis revealed that introduction of the Aid transgene in TCL1 mice resulted in mutations that corresponded closely with the AID mutational signatures in the progressive CLL patients (Figure 6A; supplemental Figure 6). A higher number of c-AID mutations, which can be directly ascribed to AID activity, were shared between the DT-AID model with progressive human disease. In total, 20 genes harboring 64 c-AID mutations in the DT-AID (lower red ticks) and 26 in the human progressive cohort (upper red ticks) were simultaneously affected (Figure 6A, largest circos plot), whereas no genes showing c-AID hotspots were affected in the TCL-1 and progressive CLL cohort (Figure 6A small circus plot). Notably, a much higher number of nonimmunoglobulin genes were concomitantly mutated in DT-AID and progressive CLL (45 c-AID, 7 nc-AID, and 19 NCG) than in TCL-1 and progressive patients (3 c-AID, 2 nc-AID, and 4 NCG) (Figure 6B).

To further explore the role of AID in CLL progression, we searched for specific nonimmunoglobulin genes carrying potentially functionally impacting mutations in the DT-AID models and human CLLs. Twenty genes were affected in DT-AID and progressive CLL; of these, 7 displayed c-AID mutations and 3 nc-AID mutations. Interestingly, Prkch, Pax-5, Mcl1, Hist1h1b, Hist1h1c, and Hist1h1d, bearing an AID signature, were identified in the DT-AID model, in the Ki-67⁺ murine fraction, and in the human progressive cohort (Figure 6C).

Altogether, these data confirmed that the mutational pattern developed by the introduction of the Aid transgene into the TCL-1 model emulates the genotype in progressive CLLs in which persistent AID expression was documented.

Of the 11 genes with nonsynonymous c-AID mutations in the Ki67⁺ fraction mentioned herein, Pax-5, Pim-1, Dusp-2, Mcl1, Gata-3, Klf2, Hist1h1e, and Hist1h1d are considered human tumor drivers.⁵⁵  Notably for Pim-1, Mcl-1, Hist1h1e, and Hist1h1d, the amino acid changes occurred at identical positions in the DT-AID mice and human neoplasms. Analyses of available crystal structures for these proteins predict that these mutations have functional consequences (Figure 7).

For PIM-1, a serine/threonine kinase overexpressed in CLL,⁶¹  Ser₉₇ is changed in both DT-AID models and CLL. This mutation is crucial in regulating protein kinase activity. Figure 7A illustrates the ionic bridge between Glu₈₉ and Lys₆₇ essential to achieve kinase activation.⁶² 

In the case of MCL-1, a key protein in lymphoid development and survival,⁶³  we identified a shared Ser→Thr replacement (murine Ser₂₆₆ and human Ser₂₈₅) (Figure 7B). Our structural analysis comparing PDB 5W89 to 5LOF⁶⁴  suggests 2 putative, not mutually exclusive scenarios. Either the Ser is important for posttranslational regulatory signaling or it plays an allosteric role in modulating the BH3-binding groove.

Concerning the histone H1 family, we exploited the available structures of H1 and H5 alone and in complex with DNA (PDBs 1GHC, 4QLC, and 5NL0).^58-60  In particular, the Ser₉₀→Arg substitution in H1.3 could lead to steric hindrance of the histone favoring chromatin decompaction (Figure 7C inset). Interestingly, this specific mutation has also been identified in CLL.⁵²  For histone H1.4, 2 specific c-AID mutations result in 2 different amino acid changes (Ala₁₆₄→Ser and Ala₁₆₇→Val). Again, both mutations have also been described in CLL.^30,52 

Remarkably, 3 of 5 of these shared mutations led to the introduction of the same amino acid in mice and humans (Figure 7) and for the other 2, the substitutions were conservative as there is high physicochemical similarity between Ser and Thr (Hist1h1e) and between polar and almost isosteric residues Thr and Asn (Pim-1) (Figure 7A,D).

Altogether, the fact that each of these mutations occurs at the same positions in the respective genes and that the changes are either identical or chemically conservative in DT-AID mice and CLL patients strongly suggests that these mutations were selected because they influence the development and/or evolution of CLL.

A long-standing model of cancer evolution posits that tumor cells progress through stages via an iterative process of expansion, diversification, and selection.⁶⁵  Clonal expansion can promote genetic diversification by the acquisition of new mutations and selection of subclones containing “advantageous” mutations that favor disease progression and/or therapeutic refractoriness.⁶⁶  Although in CLL cells microenvironmental influences increase AID expression and drive CLL progression,^17,18,23,67  the association of AID activity, disease evolution, and poor clinical outcome in progressive CLL patients remains hypothetical due to a lack of direct experimental evidence that specific off-target mutations in CLL clones are involved in disease progression.

Therefore, to better understand how AID overexpression in CLL patients promotes disease progression, we developed 2 murine strains overexpressing AID in the TCL1 background. Consistent with our hypothesis, both DT-AID models experienced more rapid and aggressive IgM⁺CD5⁺ CLL-like disease, resembling the worse clinical outcomes of patients with U-CLL expressing AID in leukemic B cells.^15,16,18,67  Because in the TCL1/act-AID mice AID is expressed globally, malignant T-cell lymphomas could arise.²¹  Nevertheless, TCL1/act-AID mice gave rise to CD5⁺ B-cell leukemia instead of a T-cell lymphoma, probably because of an earlier temporal expansion of CD5⁺ clones (supplemental Figure 4). This effect could be due to the antiapoptotic actions of TCL1 in murine B lymphocytes.⁶⁸ 

The leukemic cells in DT-AID mice grew faster and led to more rapid death. Because of this and AID’s expression being increased in proliferating human subsets of the leukemic clone,^18,23,69  we performed WES not only of the whole leukemic clone but also of leukemia cells stratified into proliferating (Ki67⁺) and nonproliferating (Ki67⁻) fractions. Because mutations bearing AID signatures affected mainly coding regions, we focused mainly on mutations with potential functional impact. Because c-AID mutations were enriched in the whole leukemic clone and in the IgM⁺CD5⁺Ki67⁺ fractions of DT-AID animals, correlation of AID activity with cell division was evident.⁴³  The findings in the Ki67⁺ fraction pinpoint key genes altered by c-AID mutations in the Wnt pathway, which are frequently mutated and involved in CLL progression.⁴⁵  Despite this interesting observation, our data in the human cohort do not identify c-AID mutations in Wnt-associated genes, possibly due to interpatient heterogeneity. A larger cohort of patients expressing AID needs to be analyzed to correlate c-AID mutations with the constitutive activation of the Wnt axis in CLL.

Moreover, genes described as AID targets and/or as gene drivers in B-cell neoplasms were only found in the proliferating fraction (Pax-5, Irf4, and Ube2a). Additionally, the genes with the highest number of c-AID mutations (Hist1h1c, d, and e, Pim1, Mcl1, Myc, and others) were shared with components of the resting fraction, indicating that they occurred upon activation of resting cells. This is consistent with cycling cells emanating from resting ones⁶⁹  and that AID is only active in cycling cells.⁴³  Hence the new, nonshared AID-induced mutations facilitate the emergence of new clonal variants that could be selected for enhanced tumor fitness.

Recently, unbiased sequencing on CLL genomes confirmed the existence of genetic heterogeneity among patients and revealed important intratumor heterogeneity.³⁰  These data are in agreement with patients differing in clinical course having different proliferation rates and ratios of proliferating and quiescent fractions expressing (or not) AID, respectively.^{4,17,23,44,69,70}  Previous analyses of the AID mutational signatures in CLL performed on nonstratified cohorts^13,31,33,71  identified AID-induced genomic uracil formation not restricted to immunoglobulin genes.^13,31,33  Additionally, analyses of an indolent CLL cohort showed a preponderance of noncanonical (20%) over canonical AID signature (5%), whereas an aging signature accounted for 75% of the total mutations.^72,73 

Here, by restricting our WES analyses to leukemic cells overexpressing AID, we documented a mutational pattern dominated by a c-AID signature in the unseparated and Ki67⁺ subsets of DT-AID clones compared with the TCL1 counterpart. Considering that TCL-1 recapitulates progressive and U-CLL,^19,20,74  and that AID overexpression associates with a clinically poor outcome in U-CLL patients,^15,67,75,76  we further compared our mouse WES findings with those of progressive U-CLLs with sustained AID expression. This revealed that overexpression of AID in the TCL-1 model drives more progressive disease with increased off-target c-AID mutations. Again, this particular pattern recapitulated many of the mutations found in progressive human CLL patients expressing AID, strongly suggesting a direct link between AID overexpression and disease progression.

Finally, we identified genes (Prkch, Pax-5, Mcl1, Hist1h1b, Hist1h1c, Hist1h1d) carrying an AID signature in DT-AID and human CLLs that could be have a role in leukemia progression after upregulation of AID enzyme.

To further explore this hypothesis, we compared genes bearing c-AID signatures in DT-AID animals to mutations described in the same genes in human tumors. Four genes (Pim-1, Mcl-1, Hist1h1d, and Hist1h1e) were mutated at homologous base pairs in mice and humans, and all occurred within a c-AID context. On the basis of available crystal structures, we modeled the sites of those mutations, providing structural bases for putative functional effects in cancer progression (Figure 7). Specifically for Moloney murine leukemia virus 1 (PIM-1) and MCL-1 proteins, the structural changes could be relevant for CLL progression because PIM kinases are essential for CLL survival^40,77  and MCL-1 levels correlate with resistance to fludarabine therapy.⁷⁸  Additionally, the MCL-1 Ser₂₈₅ replacement could allosterically modulate the BH3-binding groove and change the groove’s affinity for antiapoptotic BH3-containing proteins, leading to therapeutic resistance to MCL-1 inhibitors (Figure 7B).

A striking observation was the quantity of c-AID mutations found in histone family genes. H1 histones are necessary for the condensation of nucleosome chains and for regulation of gene transcription through chromatin remodeling and DNA methylation.⁷⁹  Interestingly, Hist1h1b, Hist1h1c, Hist1h1d, and Hist1h1e, which are drivers in human CLL,^30,78  in DLBCL⁵³  and in FL,⁸⁰  are mutated in the AID-DT models as well as in the human progressive disease.

Interestingly, the sites of c-AID mutations observed in Hist1h1e and Hist1h1d are identical between mouse and human and in some cases even the amino acid changes are the same. Hence, we propose a feed-forward mechanism whereby c-AID mutations affecting the H1 histone family favor the formation of single-stranded DNA, which in turn allows further mutagenesis by AID. Supporting this hypothesis is the Ser₉₀→Asn substitution in Hist1h1d that could lead to physical clash of the histone with the nucleosome DNA promoting DNA decompaction (Figure 7C).

It is noteworthy that point mutations in PIM-1, MCL-1, H1.3, and H1.4 occurred at the same homologous positions in our murine models and a published human CLL cohort (Figure 7).⁵²  Even more remarkable is that among 5 shared mutations, 3 led to the same and the other 2 to conservative amino acid substitutions based on physiochemical similarity. The chances of finding mutations leading to identical or biochemically equivalent amino acid changes in driver genes in 2 different unrelated species developing the same malignancy is highly unlikely to be random, and strongly suggests that a selection of these changes confers functional advantages for leukemia progression.

In summary, we describe 2 novel transgenic mouse models that recapitulate aggressive CLL and are useful tools to study AID off-target mutations in genes related with disease progression in a leukemic context. We also describe for the first time the mutational pattern of a progressive U-CLL cohort overexpressing AID and confirm that nonimmunoglobulin mutations with a c-AID signature found in the DT-AID models occur in these progressive U-CLL clones. These findings allowed us to highlight previously defined CLL mutations in nonimmunoglobulin genes, and identify new ones, some of which are tumor drivers in human B-cell leukemias/lymphomas and could therefore be relevant for the disease. Collectively, these new findings support a direct association between an activated leukemic B-cell population, aberrant AID action, and clonal aggressiveness in CLL.

The authors thank Davide F. Robbiani and Michel C. Nussenzweig for sharing the Igk-AID Tg mice, Hossein Khiabanian for analytic suggestions, Gimena Dos Santos for helpfulness assistance, and Diego Alvarez from Centro Austral de Tecnología Genómica (catg.cl) for his expert bioinformatics advice.

This work was supported, in part, by grants from the Fondo para la Convergencia Estructural del Mercosur (COF 03/11, CSIC I+D_2014, and FCE_1_2011_1_7273 (P.O.); a Consejo Nacional de Investigaciones Científicas y Técnicas (CONICET) postdoctoral fellowship and funding from the Agencia Nacional de Investigación e Innovación (ANNI) and Televie (FNRS 7.8506.19) (P.E.M.); Fondo Nacional de Desarrollo Científico y Tecnológico (FONDECYT) grants 1180882 and MAG1895 (M.A.N.); and contributions from The Nash Family Foundation, The Marks Foundation, the Karches Family, and the Jean Walton Fund for Leukemia, Lymphoma, and Myeloma Research (N.C. and K.R.R.).

Contribution: M.A.N., N.C., and P.O. conceived and designed the study; M.A.N., K.R.R., N.C., and P.O. provided study materials; V.R. provided patient samples; P.E.M., X.-J.Y., M.E.M., T.B., G.F.-G., N. Seija, N. Sotelo, C.A., M.C., N.R., and S.P.M. conducted experiments; J.S. and M.A.N. performed bioinformatics analysis; P.E.M., X.-J.Y., M.E.M., F.P., N.R., A.B., M.A.N., and P.O. analyzed and interpreted data; P.E.M., M.A.N., J.M.D.N., A.B., N.C., and P.O. wrote the manuscript; and all authors read and approved the final manuscript

Conflict-of-interest disclosure: P.O. is founder and chief hematology-oncology scientific officer of Ardan Immuno Pharma. The remaining authors declare no competing financial interests.

Correspondence: Pablo Oppezzo, Institut Pasteur de Montevideo, Research Laboratory on Chronic, Lymphocytic Leukemia, Montevideo 11400, Uruguay; e-mail: poppezzo@pasteur.edu.uy; Nicholas Chiorazzi, Karches Center for Oncology Research, The Feinstein Institutes for Medical Research, Manhasset, NY 11030; e-mail: nchizzi@northwell.edu; and Marcelo A.` Navarrete, School of Medicine, University of Magallanes, Av Bulnes 01855, 6210427 Punta Arenas, Chile; e-mail: marcelo.navarrete@umag.cl.