Mouse Models in the Study of Mature B-Cell Malignancies

Abstract

Over the past two decades, genomic analyses of several B-cell lymphoma entities have identified a large number of genes that are recurrently mutated, suggesting that their aberrant function promotes lymphomagenesis. For many of those genes, the specific role in normal B-cell development is unknown; moreover, whether and how their deregulated activity contributes to lymphoma initiation and/or maintenance is often difficult to determine. Genetically engineered mouse models that faithfully mimic lymphoma-associated genetic alterations represent valuable tools for elucidating the pathogenic roles of candidate oncogenes and tumor suppressors in vivo, as well as for the preclinical testing of novel therapeutic principles in an intact microenvironment. Here we summarize what has been learned about the mechanisms of oncogenic transformation from accurately modeling the most common and well-characterized genetic alterations identified in mature B-cell malignancies. This information is expected to guide the design of improved molecular diagnostics and mechanism-based therapeutic approaches for these diseases.

B-cell lymphomas comprise a heterogeneous group of neoplasms that originate from the oncogenic transformation of cells at various stages of mature B-cell development and, in the vast majority of cases, from germinal center (GC) B cells (Stevenson et al. 1998; Küppers et al. 1999; Swerdlow et al. 2016). A common theme that emerged from decades of studies aimed at the identification of lymphoma-associated genetic alterations is the recurrent involvement of genes that play critical regulatory roles at specific B-cell developmental stages, resulting in their aberrant expression or function and strongly implicating a mechanistic role for these proteins in lymphoma pathogenesis (Shaffer et al. 2012; Pasqualucci 2019). However, the molecular mechanisms by which these alterations contribute to lymphoma initiation and progression are often difficult to investigate, because suitable in vitro model systems that mimic the biological complexity of antigen-dependent B-cell development, and in particular the GC reaction, are lacking. The mouse is a versatile model organism for studying higher vertebrates, because its immune system has the sophistication of the human counterpart and mouse embryonic stem cells are amenable to directed genetic manipulation, allowing to mimic the structural aberrations identified in cancer. As such, genetically engineered mouse models (GEMMs) have been a mainstay of cancer biology in the effort to determine the pathogenic roles of candidate oncogenes and tumor suppressors in vivo. Additionally, the availability of accurate animal models provides a valuable opportunity for preclinical studies aimed at evaluating the expanding armamentarium of molecularly targeted drugs. Of course, a number of caveats must be considered when approaching these studies, such as potential species-specific biological differences in the target cell of the oncogenic transformation process. Nevertheless, modeling lymphoma-associated genetic lesions in mice has offered critical insights into the mechanisms underlying the tumorigenic process and will continue to be a powerful filter for the nomination of potential cancer drivers among the numerous novel candidates uncovered by recent genome-sequencing studies. This review provides an examination of the advances obtained from the faithful modeling of lymphoma-associated genetic lesions in transgenic mice, as related to the molecular pathogenesis of the most common B lymphoid neoplasms: mantle cell lymphoma (MCL), Burkitt lymphoma (BL), follicular lymphoma (FL), diffuse large B-cell lymphoma (DLBCL), and chronic lymphocytic leukemia/small lymphocytic lymphoma (CLL/SLL). Emphasis will be placed on GEMMs that were proven to accurately recapitulate the human disease both genetically and phenotypically. We refer to other reviews for an in-depth discussion of alternative in vivo experimental systems that could serve as a more rapid approach to enable gain- and loss-of-function studies or the preclinical testing of novel therapeutic concepts, including the adoptive transfer of genetically modified hematopoietic progenitor cells (HPCs) (Oricchio et al. 2010) and patient-derived tumor xenograft mouse models (Townsend et al. 2016; Pizzi and Inghirami 2017).

MATURE B-CELL DEVELOPMENT

During normal B-cell development, three B-cell lineages with specific biological functions can be distinguished in higher vertebrates: CD5⁺ B cells, marginal zone B cells, and follicular (naive) B cells. B cells expressing the CD5 antigen predominate early in life, although they continue to be generated throughout ontogeny (Baumgarth 2017). They produce polyreactive antibodies against common bacterial pathogens and provide a first line of defense against invading microorganisms. Marginal zone B cells are located at the entry site of blood-borne pathogens in the spleen, where they quickly respond to antigenic challenge by secreting antibodies that clear the pathogen (Lopes-Carvalho and Kearney 2004). Finally, in T-cell-dependent, adaptive immune responses, naive B cells undergo antibody maturation to exogenous antigens, which ultimately results in the generation of plasma cells capable of secreting high-affinity antibodies against the invading pathogen, as well as of memory B cells that quickly respond in recall responses against the same antigen (Rajewsky 1996). A key step in this process is the recruitment of naive B cells into the GC reaction following antigen encounter and guidance provided by antigen-specific T cells (Fig. 1; Victora and Nussenzweig 2012; De Silva and Klein 2015; Mesin et al. 2016; Bannard and Cyster 2017). Within the GC microenvironment, B cells first differentiate into centroblasts or dark zone (DZ) B cells. These cells proliferate at a high rate and modify their rearranged immunoglobulin variable (IgV) region genes through the process of activation-induced cytidine deaminase (AID)-mediated somatic hypermutation (SHM) to generate antibody specificities with improved affinity to the pathogen. Centroblasts then move to a different area within the GC, known as the light zone (LZ), where they cease proliferating and differentiate into centrocytes or LZ B cells before being selected for improved antigen binding with the help of T-follicular helper (Tfh) cells. GC B cells that are not rescued by Tfh cells because the newly introduced somatic mutations led to a decrease in affinity or disrupted the antibody structure are destined to undergo apoptosis. Positively selected LZ B cells can recycle to the DZ to undergo further rounds of SHM and selection and eventually differentiate into plasma cells or memory B cells, which constitute the immunological memory in the antibody system (Fig. 1). The GC LZ was also thought to represent the site where B cells undergo class switch recombination (CSR), a second AID-dependent B-cell-specific DNA-modification process that leads to the generation of antibodies with different effector functions but identical specificities; however, recent work provided compelling evidence that this reaction takes place predominantly prior to the GC reaction and to SHM (Roco et al. 2019). Of note, single-cell analysis of gene expression and somatically mutated IgV region genes in human GC B cells revealed a continuum of transcriptional changes as cells bidirectionally recirculate between the DZ and LZ compartment (Milpied et al. 2018).

Figure 1.

Presumed cellular origin of common mature B-cell malignancies for which bona fide genetically engineered mouse models (GEMMs) are available. Relationship between distinct types of B-cell neoplasms and their presumed normal counterpart, as determined based on phenotypic and genetic similarities. The major B-cell developmental steps following antigen encounter are indicated, along with CD5⁺ mature B cells. Only GEMMs are shown that were obtained by recapitulating genetic changes observed in the human disease and that develop tumors featuring key characteristics of the corresponding human lymphoma entity. See text for details. B_DZ, dark zone B cells; B_LZ, light zone B cells; Tfh, T-follicular helper cell; CLL, chronic lymphocytic leukemia; BL, Burkitt lymphoma; FL, follicular lymphoma; GCB-DLBCL, germinal center B-cell-like diffuse large B-cell lymphoma; ABC-DLBCL, activated B-cell-like diffuse large B-cell lymphoma.

CELLULAR COUNTERPARTS OF B-CELL LYMPHOMAS

A correct understanding of the putative normal counterpart of distinct lymphoid neoplasms is critical for the construction of faithful mouse models of the disease, because the genetic lesion of interest should be targeted to the proper cellular context both temporally and spatially (Fig. 1).

As an irreversible marker of transit through the GC, the determination of the IgV region gene SHM status, in combination with morphological and immunophenotypic analysis of normal and malignant B cells, has long provided information on the putative normal cellular counterparts of various lymphoma types (Stevenson et al. 1998; Küppers et al. 1999). The advent of genome-wide gene expression profiling technologies furthered these notions, allowing a more refined assignment of diverse lymphoma entities to their presumptive normal counterpart, as well as the identification of functionally relevant disease subtypes (Shaffer et al. 2006; Klein and Dalla-Favera 2008).

Early studies established that, with the exception of most MCL cases, the majority of B-cell lymphoid malignancies, including BL, FL, DLBCL, Hodgkin lymphoma, and possibly marginal zone lymphoma (not covered here), derive from a GC-experienced B cell, as evidenced by the expression of somatically hypermutated IgV region genes (Stevenson et al. 1998; Küppers et al. 1999). The cellular origin of CLL/SLL, where two major subtypes can be recognized based on the presence or absence of mutations in the IgV region genes, has long remained elusive and is in part still being debated (Chiorazzi and Ferrarini 2011). Nonetheless, multiple lines of evidence, including a shared gene expression profile, suggest that both CLL subtypes derive from antigen-experienced B cells (Klein et al. 2001; Rosenwald et al. 2001; Kipps et al. 2017). Of note, transcriptomic analysis has drawn a possible relationship with a mature CD5-expressing B-cell subpopulation in adults that comprises ∼1% of peripheral blood B cells, expresses the CD27 memory B cell marker, and displays both mutated and unmutated IgV region genes (Seifert et al. 2012).

MCL cases can be distinguished by a gene and epigenetic signature that identifies conventional and leukemic nonnodal MCL, with the latter showing a more favorable outcome (Puente et al. 2018). The cellular counterpart of conventional MCL, the most common form, appears to be a naive B cell with no or few somatic mutations in its IgV region genes, whereas the nonnodal MCL type shows memory B-cell-like features (Clot et al. 2018).

In the case of DLBCL, at least two distinct phenotypic subtypes have been recognized based on their similarity to GC B cells (the so-called GCB-DLBCL) or in vitro activated B cells (ABC-DLBCL), with ∼20% of cases remaining unclassified (Alizadeh et al. 2000). Although further subdivision within this heterogeneous disease has been revealed by genetic profiling (Chapuy et al. 2018; Schmitz et al. 2018), the historic classification into GCB-DLBCL and ABC-DLBCL has proved fundamental in the recognition of distinct normal cellular counterparts for these two tumor types. More recently, the identification of cell surface markers that distinguish DZ and LZ GC B cells—namely, CXCR4 on DZ B cells and CD86 on LZ B cells—allowed the definition of comprehensive gene expression signatures for these subpopulations (Victora et al. 2010) and a more precise assignment of BL, FL, and DLBCL to putative normal cellular counterparts (Victora et al. 2012). Perhaps surprisingly, whereas BL shows a gene expression profile that is closely related to DZ B cells, FL and GCB-DLBCL are transcriptionally more similar to LZ B cells, possibly representing an intermediate GC B-cell stage (Victora et al. 2012). The ABC-DLBCL, on the other hand, may correspond in vivo to a small subset of LZ B cells poised to undergo differentiation into plasma cells and could also include, as recently suggested, cases with similarities to memory B cells (unpublished observations) or to extrafollicular marginal zone B cells (Chapuy et al. 2018; Schmitz et al. 2018).

Although many models have been generated for the overexpression or deletion of oncogenes and tumor-suppressor genes (Ramezani-Rad and Rickert 2017), some of the strategies would not recapitulate genetic alterations found in the human disease or the exact timing of the genetic lesion, which may impact on the interpretation of the results. A major advance in this regard came from the use of Cre transgenes that can be expressed at various B-cell differentiation stages through developmentally controlled promoters, allowing them to specifically and inducibly activate or inactivate the gene of interest in the desired cell type. Thus, the crossing of floxed alleles to mb1-Cre (Hobeika et al. 2006), CD19-Cre (Rickert et al. 1997), and CD21-Cre mice (Kraus et al. 2004) allows gene recombination in B cells at a particular early B-cell developmental phase, throughout their development, or specifically in mature B cells, respectively. Cγ1-Cre and AID-Cre mouse strains are used to induce gene recombination specifically in antigen-activated B cells, including GC B cells (Casola et al. 2006; Crouch et al. 2007). Figure 2 illustrates the strategies utilized for the generation of mouse models discussed in this review.

Figure 2.

Strategies used for the generation of genetically engineered mouse models (GEMMs) of B-cell malignancies. Simplified schematic representation of the targeting alleles and experimental strategies used to reconstruct genetic alterations associated with human B-cell lymphomas in GEMMs (alone or in combination with additional oncogenes). The adoptive transfer approach is also illustrated as a tool to introduce conditional or inducible genetically manipulated alleles. See text for details. Only representative mouse models are listed (see Table 1 for a full list).

Table 1.

Genetically engineered mouse models (GEMMs) of mature B-cell malignancies

Gene	Disease	Mutation type	Mouse model	Approach	Target cell	Phenotype	Reference(s)
Myc/PI3K	BL	Deregulated expression + gain-of-function mutation	R26StopFLMyc;R26StopFLP100*; Cγ1- Cre	Conditional KI	GC B cells	BL	Sander et al. 2012
Bcl2	FL, GCB-DLBCL	Deregulated expression	VavP-Bcl2 Bcl2-Ig BCL2trac	Transgene insertion Transgene insertion TI/adoptive transfer	HPC B cells B cells	FL FL FL	Egle et al. 2004 McDonnell et al. 1989 Sungalee et al. 2014
Kmt2dc	FL, DLBCL	Genetic deletion	Kmt2dfl/fl;VavP-Bcl2;Cγ1-Cre Kmt2dfl/fl;VavP-Bcl2;Cd19-Cre	Conditional KO Conditional KO	GC B cells B cells	FL, DLBCL FL, DLBCL	Zhang et al. 2015 Zhang et al. 2015
Crebbpc	FL, DLBCL	Genetic deletion	Crebbpfl/+;VavP-Bcl2;Cγ1-Cre Crebbpfl/fl;VavP-Bcl2;Cd19-Cre Crebbpfl/fl;Eμ-Bcl2; Mb1-Cre	Conditional KO Conditional KO Conditional KO	GC B cells B cells Early B cells	FL FL FL, DLBCL	Zhang et al. 2017 Zhang et al. 2017 García Ramírez et al. 2017
Ezh2c	FL, GCB-DLBCL	Gain-of-function mutation	Ezh2Y641F/+;IμHABCL6;Cγ1-Cre Ezh2Y641F/N;VavP-Bcl2	cKI/adoptive transfer Adoptive transfer	GC B cells GC B cells	DLBCL FL, DLBCL	Béguelin et al. 2016 Béguelin et al. 2013; Ennishi et al. 2019b
Mef2b	FL, DLBCL	Gain-of-function mutation	Mef2bD83V/+;CD21-Cre Mef2bD83V/+;BCL2-Ig;CD21-Cre	Conditional KI Conditional KI	GC B cellsa GC B cellsa	FL, DLBCL FL, DLBCL	Brescia et al. 2018 Brescia et al. 2018
Rragc	FL	Gain-of-function mutation	RragcS74C/+ or RragcT89N/+; VavP-Bcl2	TI/adoptive transfer	All cells	FL	Ortega-Molina et al. 2019
Gna13	GCB-DLBCL, BL	Genetic deletion	Gna13fl/fl;Mb1-Cre Gna13fl/fl;R26StopFLMyc;AID-Cre	cKO/adoptive transfer Conditional KO	GC B cellsb GC B cells	GC BCL GC BCL	Muppidi et al. 2014 Healy et al. 2016
Bcl6	DLBCL	Deregulated expression	IμHABCL6	KI	GC B cells	DLBCL	Cattoretti et al. 2009
Prdm1	ABC-DLBCL	Genetic deletion	Blimp1fl/fl;Cγ1-Cre Blimp1fl/fl;R26StopFLIkk2ca; Cγ1-Cre	Conditional KO Conditional KO	GC B cells GC B cells	DLBCL DLBCL	Mandelbaum et al. 2010 Calado et al. 2010
Myd88	ABC-DLBCL	Gain-of-function mutation	Myd88p.L252P/+;Cd19-Cre	Conditional KI	B cells	LPD, DLBCL	Knittel et al. 2016
Tet2	FL, GCB-DLBCL	Genetic deletion	Tet2fl/fl;Vav-Cre Tet2fl/fl;IμHABCL6;Cγ1-Cre	Conditional KO cKO/adoptive transfer	HPC GC B cells	BCL BCL	Dominguez et al. 2018 Dominguez et al. 2018
miR-15a/16-1	CLL	Genetic deletion	mir-15a/16-1fl/fl;Cd19-Cre mir-15a/16-1−/−	Conditional KO KO	B cells All cells	CLL (DLBCL) CLL (DLBCL)	Klein et al. 2010 Klein et al. 2010
13q14-MDR	CLL	Genetic deletion	13q14-MDRfl/fl;Cd19-Cre 13q14-MDR−/−	Conditional KO KO	B cells All cells	CLL (DLBCL) CLL (DLBCL)	Klein et al. 2010 Klein et al. 2010
13q14-CDR	CLL	Genetic deletion	13q14-CDRfl/fl;Cd19-Cre 13q14-CDR+/−	Conditional KO KO	B cells All cells	CLL (DLBCL) CLL (DLBCL)	Lia et al. 2012 Lia et al. 2012
Sf3b1d	CLL	Gain-of-function mutation	Sf3b1K700L/+;Atmfl/fl;Cd19-Cre	Conditional KI	B cells	CLL	Yin et al. 2019
Irf4d	CLL	Genetic deletion	Irf4−/−;Vh11	KO	All cells	CLL	Shukla et al. 2013

Listed are only GEMMs obtained by reconstructing genetic lesions associated with human B-cell malignancies and recapitulating key characteristics of the disease. See text for mouse models generated by adoptive transfer of transduced HPCs.

BL, Burkitt lymphoma; FL, follicular lymphoma; BCL, B-cell lymphoma; GCB-DLBCL, germinal center B-cell-like diffuse large B-cell lymphoma; ABC-DLBCL, BCL, B cell lymphoma activated B-cell-like diffuse large B-cell lymphoma; CLL, chronic lymphocytic lymphoma; KI, knock-in; KO, knockout; cKI, conditional knock-in; cKO, conditional knockout; TI, targeted insertion; GC, germinal center; HPC, hematopoietic progenitor cell.

^aCre-mediated recombination in mature B cells, but the endogenous promoter is activated in GC B cells.

^bCre-mediated recombination in all B cells, but the endogenous promoter is activated or up-regulated in GC B cells.

^cGC-derived lymphomas observed only in cooperation with Bcl2 deregulated expression.

^dLymphomagenesis observed only in cooperation with Atm deletion (Sf3b1) or Vh11 (Irf4).

MANTLE CELL LYMPHOMA

MCL represents 5% of all lymphoma diagnoses and comprises two recognized biological entities: an aggressive subtype characterized by unmutated or minimally mutated IgV region genes (conventional MCL) and a more indolent subtype that carries IgV mutations and is frequently leukemic (nonnodal MCL) (Jares et al. 2012; Swerdlow et al. 2016; Puente et al. 2018). The genetic hallmark of both diseases is the t(11;14) translocation that juxtaposes the CCND1 gene (formerly known as BCL1), encoding for the cell-cycle regulatory protein cyclin D1 and normally not detected in resting B cells, to enhancer sequences in the Ig locus, leading to its constitutive expression and to loss of cell-cycle control (Jares et al. 2012).

The pathogenic role of cyclin D1 deregulation in human neoplasia is supported by the ability of the overexpressed protein to accelerate B-cell lymphomagenesis in transgenic mice, when combined with other oncogenic events such as MYC deregulation (Bodrug et al. 1994; Lovec et al. 1994). However, the type of lymphomas produced in Eµ^CycD1/Eμ-Myc double transgenic mice resembles that found in Eμ-Myc mice (i.e., mainly pre-B-cell lymphomas; see next section). Moreover, MYC overexpression and/or rearrangement is rare in MCL, occurring only in the blastoid variant. Lymphomas with the classic morphologic and phenotypic (CD5⁺CD23⁻) appearance of the human conventional MCL were obtained by combining cyclin D1 overexpression with homozygous loss of the pro-apoptotic gene Bim1 in Eµ^CycD1/CD19^Cre/Bim1^fl/fl mice, although at very low frequencies unless challenged with serial antigenic stimulation (Katz et al. 2014). Finally, transgenic mice with a nuclear export-deficient CCND1 (D1/T286A mice) developed lymphomas by 18 mo of age with 50% penetrance, but these tumors were derived from early mature B cells, according to their IgM^hiIgD^lo immunophenotype (Gladden et al. 2006). Thus, an animal model for bona fide MCL is still lacking.

A blastoid variant of MCL has been described that frequently harbors deletions of p16^INK4a, an inhibitor of the cyclin-dependent kinase CDK4, and displays overexpression of both CDK4 and MYC (Beà et al. 1999). Crossing a knock-in strain that carries a mutant CDK4 protein resistant to inhibition by p16^INK4a with Myc transgenic mice (Cdk4^R24C/Myc-3′RR mice) leads to lymphomas that phenotypically resemble human blastoid MCL (Truffinet et al. 2007; Vincent-Fabert et al. 2012). Interestingly, a distinctive feature of these tumors is the accumulation of high amounts of CCND1/CDK4 complexes, even though this mouse model does not carry a Ccnd1 transgene. Thus, although not recapitulating the precise genetic aberrations observed in human MCL, this system was useful in ascertaining the tumor-promoting role of genes known to be involved in MCL pathogenesis.

Recent exome-sequencing studies have provided a comprehensive picture of the coding mutations found in this lymphoma type, identifying additional candidate genes with potential roles in MCL pathogenesis (Beà et al. 2013). It is expected that such improved knowledge will lead to the development of murine MCL models more closely resembling the human disease.

BURKITT LYMPHOMA

BL, an aggressive B-cell lymphoma, is invariably associated with chromosomal translocations that bring the MYC gene under the control of one of the Ig enhancers, causing its ectopic expression in the bulk GC population in which this protein is otherwise not present (Calado et al. 2012; Dominguez-Sola et al. 2012). MYC translocations are found in all three clinical variants recognized within this entity by the WHO classification: (1) endemic BL (eBL), which occurs in equatorial areas (sub-Saharan Africa and South America) and is frequently associated with endemic malaria; (2) sporadic BL (sBL), which occurs elsewhere; and (3) immunodeficiency-associated BL, which most commonly occurs in HIV-positive individuals developing acquired immune deficiency syndrome (AIDS-BL). Another characteristic of BL is the frequent association with Epstein–Barr virus (EBV). The morphological features and the expression of somatically mutated IgV region genes in all BL cases have long identified a GC B cell as the normal cellular counterpart of this disease (Stevenson et al. 1998; Küppers et al. 1999). Moreover, the evidence of ongoing SHM in at least a fraction of cases indicated a cellular origin from DZ B cells/centroblasts that are actively undergoing SHM; this notion found support in the observation that the transcriptional profile of BL is closely related to the GC DZ signature (Victora et al. 2012). Recent whole-exome sequencing (WES) studies identified additional recurrently targeted genes in this lymphoma, including TCF3, which encodes the transcription factor E2A and harbors gain-of-function mutations in 10%–25% of cases, and ID3, a negative regulator of E2F that is typically affected by inactivating mutations in 35%–58% of all BL subtypes (Love et al. 2012; Richter et al. 2012; Schmitz et al. 2012). The TCF3-ID3 axis is predicted to promote tonic B-cell receptor (BCR) signaling, leading to the aberrant activation of the phosphoinositide-3-kinase (PI3K) signaling pathway, which is normally not active in DZ B cells. Moreover, a subset of eBL cases shows gain-of-function mutations of CCND3, a D-type cyclin required for GC formation and the proliferation of DZ B cells (Peled et al. 2010; Cato et al. 2011; Richter et al. 2012; Schmitz et al. 2012). Finally, the FOXO1 transcription factor is a frequent target of missense mutations, many of which cluster around the known T24 phosphorylation site (12% of cases) (Schmitz et al. 2012). Studies in transfected cell lines have shown that mutations in FOXO1 prevent its cytoplasmic translocation and functional inactivation, which is the physiological consequence of PI3K signaling (Trinh et al. 2013). However, a mechanistic basis for the aberrant nuclear localization of FOXO1, observed in virtually all BL cases independent of genetic alterations, is presently lacking.

Because of the identification of MYC as the first translocated proto-oncogene in B-cell lymphomas, several transgenic mouse models were generated in the 1980s to 2000s (Adams et al. 1985; Bützler et al. 1997; Kovalchuk et al. 2000). In these mice, the expression of MYC transgenes was controlled by intronic heavy chain and light chain enhancers, which are active from the early stages of B-cell development but are not linked to the translocated MYC allele in most sporadic BLs. These mouse models all develop clonal B-cell proliferations at high penetrance, with some differences in their spectrum among the various strains, but generally unrelated to the normal counterpart of the human tumors and devoid of its pathognomonic phenotypic characteristics. For example, Eμ-Myc transgenic mice develop predominantly pre-B-cell lymphomas/leukemias that lack surface Ig expression (Adams et al. 1985; Harris et al. 1988). Lymphomas arising in λMYC transgenic mice are of mature B-cell origin (IgM⁺), but retain the CD43 marker and do not express somatically mutated IgV region genes (Kovalchuk et al. 2000), indicating that the target cell of transformation in these tumors is a transitional/pre-GC B cell rather than a GC DZ B cell. Analogous phenotypes were reported when different Ig enhancers were utilized to drive MYC overexpression throughout B-cell development (Park et al. 2005; Wang and Boxer 2005). As such, these mouse models have been fundamental for mechanistic studies aimed at investigating the pathogenic role of MYC in the transformation process or the cooperativity among oncogenes and the sensitivity to drugs (Hemann et al. 2003; Pasqualucci et al. 2008; Varano et al. 2017; Ravà et al. 2018). However, these models do not faithfully recapitulate key phenotypic and genetic features of human BL—or DLBCL, in which MYC can also be deregulated by chromosomal translocations (Chong et al. 2018)—and are not considered informative toward the dissection of the MYC oncogenic function in the pathogenesis of these cancers nor in the understanding of the cellular networks that are perturbed within the unique GC environment to license lymphomagenesis when MYC is ectopically expressed.

The generation of a BL-like lymphoma in mice was achieved by crossing two separate inducible transgenes that lead to MYC overexpression specifically in GC B cells, along with a mutant P110* allele mimicking a constitutively active form of PI3K (Sander et al. 2012). Tumors developing in these mice closely resemble the human BL in its typical morphologic and histologic features, gene expression profile (including the expression of BCL6), and the acquisition of tertiary transforming events, such as mutations in CCND3. Together, the shared phenotypic and genetic features between human BLs and the BL-like tumors modeled in these mice identify the Myc/P110* animal model as a valuable system for both mechanistic and preclinical studies. Moreover, tumors developing in this strain were instrumental in identifying a pro-proliferative and anti-apoptotic function of FOXO1 in the transformation of GC B cells toward BL (Kabrani et al. 2018).

FOLLICULAR LYMPHOMA

FL, the second most common type of B-cell lymphoma and the most frequent indolent lymphoma, is an incurable malignancy characterized by a continuous pattern of remissions and relapses that are accompanied by progressively reduced response to therapy and, in as many as 40% of cases, histologic transformation to a more aggressive disease (typically a DLBCL, or transformed FL [tFL]) (Lossos and Gascoyne 2011; Casulo et al. 2015). The genetic hallmark of FL is the t(14;18) translocation, a by-product of the VDJ recombination process that places the BCL2 coding domain under the control of the Ig heavy chain enhancer, leading to ectopic and constitutive expression of the antiapoptotic protein BCL2 (Kridel et al. 2012). More recently, WES efforts have uncovered highly recurrent somatic mutations in several genes encoding for histone-modifying enzymes, such as the KMT2D methyltransferase (70% of cases), the CREBBP acetyltransferase (65% of cases), and the EZH2 methyltransferase (22% of cases), as well as multiple linker-histone family members (>44% of cases), pointing to altered epigenetic regulation as a central driving force of this malignancy (Morin et al. 2010, 2011; Pasqualucci et al. 2011; Li et al. 2014; Okosun et al. 2014). Other prevalent alterations that have been successfully modeled in the mouse include gain-of-function mutations of MEF2B (15% of cases) (Ying et al. 2013), loss-of-function biallelic mutations and deletions of the TNFRSF14 gene (up to 40% of cases) (Cheung et al. 2010; Okosun et al. 2014), and point mutations of the RRAGC gene that, along with other genes involved in the activation of the mammalian target of rapamycin complex 1 (mTORC1) signaling, accounts for 17% of cases (Okosun et al. 2016). Of note, mutations in histone/chromatin modification genes, MEF2B, and TNFRSF14 are highly frequent in both FL and DLBCL and should thus be considered reflective of both diseases. However, we discuss them in this section because of their higher prevalence in FL compared to DLBCL and the preferential development of FL-like diseases in these models.

Mouse Models Recapitulating the BCL2 Translocation

Although several GEMMs have been generated to study the consequences of deregulated BCL2 expression dictated by the t(14;18) translocation, only two of them were shown to develop FL-like tumors within their lifespan, indicating that the translocation alone is insufficient to drive full malignant transformation, and that additional events are required: the VavP-Bcl2 and the BCL2-Ig mouse model (McDonnell et al. 1989; McDonnell and Korsmeyer 1991; Egle et al. 2004). Eµ-BCL2 mice exhibit an expanded small B-lymphocyte population with no spontaneous tumor formation (Strasser et al. 1991); however, they have been useful to document the cooperativity between candidate oncogenic events (García-Ramírez et al. 2017). An additional mouse model (the mosaic BCL2 tracer) is limited to the development of in situ FL, but it is worth mentioning here because it elegantly recapitulates the initial events underpinning the accumulation and expansion of BCL2-translocated B cells (Sungalee et al. 2014), which can be found in >50% of healthy subjects, even though most of these individuals will never develop the disease (Roulland et al. 2006).

In the VavP-Bcl2 mouse model, the Vav promoter drives expression of the BCL2 oncogene specifically in the hematopoietic compartment, hence at a stage prior to the occurrence of the BCL2 translocation (Ogilvy et al. 1999); nonetheless, these mice develop B-cell lymphomas that accurately recapitulate critical aspects of the genetics, pathology, and GC origin of human FLs, including a follicular pattern, the expression of peanut agglutinin and BCL6 in the absence of post-GC markers, and the presence of clonally rearranged IgV region genes that are somatically mutated (Egle et al. 2004). These tumors require the presence of T cells, which also overexpress the BCL2 transgene, because the GC hyperplasia could be abolished by antibodies directed against CD4 in vivo. Owing to its close resemblance to the human disease, the VavP-Bcl2 model has been extensively used as an experimental system to assess the cooperative activity of genetic alterations that are concomitantly mutated with BCL2 translocations, either by crossing this line with other GEMMs (e.g., Crebbp^fl/fl, Kmt2d^fl/fl, Ezh2^Y641N) or as a source of HPCs that can be transplanted into irradiated mice after retroviral transduction with constructs for gain- and loss-of-function mutants (Oricchio et al. 2014).

Different from VavP-Bcl2 mice, expression of a BCL2 minigene in the BCL2-Ig mouse model is controlled by Ig regulatory elements and is thus restricted to B cells (McDonnell et al. 1989). Similar to Eµ-BCL2 mice, this strain displays an excess of B-lineage cells (both small B cells and plasma cells) exhibiting prolonged survival in vitro and, although no tumors were detected upon 12-mo follow-up in the original studies, was later shown to accumulate GC B cells and to develop PAX5⁺BCL6⁺ FLs in as many as 40% of cases when challenged by chronic antigenic stimulation, with a smaller fraction of mice presenting PAX5⁻BCL6⁻IRF4⁺ plasmacytoid tumors (Brescia et al. 2018).

In the BCL2 tracer mouse model, expression of a functional human BCL2 transgene is contingent on its inversion by the V(D)J recombination process after adoptive transfer of progenitor cells into irradiated recipient wild-type mice (Fig. 2; Sungalee et al. 2014). This model mimics several aspects of FL pathogenesis as currently understood, including the sporadic nature of the t(14;18) and its induction at the appropriate developmental stage—that is, a BM pro-/pre-B cell, as a by-product of VDJ recombination (Küppers and Dalla-Favera 2001). Moreover, tracking of the BCL2-overexpressing B cells is possible in these mice by means of the unique coding joint generated by the recombination, together with the presence of an EYFP reporter and the use of specific antibodies against the human BCL2 protein. Investigation of the T-cell-dependent immune response in this model revealed that BCL2-overexpressing B cells require multiple GC reentries before disseminating and progressing to advanced preneoplastic stages of FL, analogous to the human FL in situ. In particular, although the fraction of BCL2⁺ cells in the naive, GC, and memory B-cell compartment was not significantly different upon a single immunization, these cells were markedly enriched in the GC and memory B-cell subset following chronic antigenic recall and were able to repopulate the GCs of immunized wild-type mice in adoptive transfer experiments (Sungalee et al. 2014). Combined with the observation that t(14;18)-positive cells in healthy individuals harbor somatically mutated IgV region genes, these data formed the basis for the current model of FL ontogenesis, which infers an origin from a recirculating memory B cell requiring multiple transits through the GC before the acquisition of additional genetic or epigenetic disturbances would ultimately drive the development of clonal tumors.

Mouse Models Recapitulating Alterations in Histone Modification Genes

Mutations in epigenetic modification genes emerged as the most common genetic event in both FL and DLBCL, collectively accounting for almost all FL cases and >50% of DLBCL cases, and revealing a significant role for epigenetic remodeling in B-cell malignancies. One important observation made from the analysis of sequential FL and tFL biopsies was that inactivating mutations of CREBBP and KMT2D represent early events in the tumor evolutionary history, which are acquired by a common mutated precursor cell (CPC) before its divergent evolution and ultimate clonal expansion to FL or tFL in the GC (Okosun et al. 2014; Pasqualucci et al. 2014; Green et al. 2015). The precise cell differentiation stage at which KMT2D and CREBBP mutations are gained is not known and could be anywhere between a hematopoietic stem cell (HSC) and the GC B cell. Therefore, attempts to reproduce the hemizygous or homozygous genetic loss of these two enzymes were made in HSCs (Horton et al. 2017), early B cells (via a CD19-Cre recombinase) (Ortega-Molina et al. 2015; Zhang et al. 2015b, 2017), and GC B cells (via the Cγ1-Cre recombinase) (Zhang et al. 2015b, 2017).

KMT2D is a catalytic component of the mammalian COMPASS (complex of proteins associated with Set1) complex that facilitates transcription through mono- and dimethylation of histone 3 lysine 4 (H3K4) at enhancer/super-enhancer regions (Sze and Shilatifard 2016). Based on its mutational spectrum that comprises mostly truncating events, KMT2D is postulated to play a central role in suppressing the lymphomagenesis process. Interestingly, conditional deletion of Kmt2d in pre-B cells, that is, much earlier than the final malignant transformation step, led to a significant expansion of the GC B-cell subpopulation compared to wild-type littermates, whereas the phenotype was less pronounced when Kmt2d was disrupted after the initiation of the GC reaction (Zhang et al. 2015b). Analogously, GC B cells from CD19-Cre compound mice showed more robust changes in their transcriptional profile compared to cells in which deletion of Kmt2d was induced in the GC (Zhang et al. 2015b). These data suggest that the timing and thus the number of cell divisions completed by the cell following KMT2D inactivation may be important in allowing the epigenetic reprogramming of the CPC. This in turn may contribute to tumor formation by creating a favorable environment for the acquisition of additional genetic or epigenetic oncogenic events, including the impaired expression of terminal differentiation programs. In line with this model, loss of Kmt2d alone in the GC was not sufficient to drive lymphomagenesis, but when combined with deregulated expression of BCL2 (as commonly seen in human FL and DLBCL) led to a significant increase in the percentage of bona fide FL and DLBCL carrying clonally rearranged, mutated IgV region genes and expressing GC-specific markers (Zhang et al. 2015b). The synergistic effect of Kmt2d loss and BCL2 deregulation in vivo was independently confirmed in a mouse model of adoptive transfer in which Kmt2d was silenced in VavP-Bcl2 HPCs prior to the reconstitution of lethally irradiated, syngeneic mice (Ortega-Molina et al. 2015).

Intriguingly, the Kmt2d-KO mouse model shares several similarities with conditional Crebbp-deficient mice, as exemplified by (i) an expanded GC population with an altered DZ:LZ ratio and partially overlapping transcriptional changes; (ii) a more prominent GC phenotype in CD19-Cre compared to Cγ1-Cre mice; and (iii) the inability to induce tumor formation as a single event, but a strong synergism with BCL2 deregulation in driving FL or, at lower frequencies, DLBCL (Zhang et al. 2017). CREBBP is a histone and non-histone acetyltransferase that modulates transcription by depositing the active H3K27 acetylation mark at promoter and enhancer domains of numerous genes, and was found to occupy virtually all GC-specific super-enhancers. However, not all CREBBP-bound genes are transcriptionally affected by its loss in purified murine GC B cells, as well as DLBCL cell lines (Jiang et al. 2017; Zhang et al. 2017), in part because of the compensatory activity of its paralog EP300 (Meyer et al. 2019). Instead, Crebbp deletion causes focal enhancer loss of H3K27Ac and reduced expression of genes that are involved in the exit from the GC reaction (Jiang et al. 2017; Zhang et al. 2017). These include downstream effectors of the BCR and NF-κB signaling pathways, several cytokines, and a number of antigen presentation molecules, the most notable being MHC class II genes (Jiang et al. 2017; Zhang et al. 2017). Indeed, decreased MHC-II expression with reduced frequency of tumor-infiltrating T-cell subsets is a distinctive feature of CREBBP-mutated human FL (Green et al. 2015). Notably, the chromatin domains occupied by CREBBP are direct targets of the BCL6 oncorepressor in a complex with SMRT and HDAC3 (Jiang et al. 2017; Zhang et al. 2017); furthermore, CREBBP can directly acetylate the BCL6 protein to inactivate its function by preventing the interaction with co-repressor complexes (Bereshchenko et al. 2002; Pasqualucci et al. 2011). Thus, a major purpose of CREBBP in the GC is to oppose the oncogenic activity of BCL6 and prime the activation of terminal differentiation/antigen presentation programs as LZ B cells engage Tfh cells and prepare to exit the GC. In line with these data, CREBBP-mutant lymphomas show reduced expression of genes that are antagonistically regulated by the BCL6-SMRT-HDAC3 complex and become dependent on HDAC3, inhibition of which restored histone acetylation at these enhancers and suppressed lymphoma growth both in vitro and in vivo (Jiang et al. 2017; Mondello et al. 2020). The identification of HDAC3 and EP300 as vulnerabilities of CREBBP-mutant cells suggests potential therapeutic avenues for these lymphomas.

EZH2 is a histone methyltransferase that is highly expressed in GC B cells and is required for GC formation (Béguelin et al. 2013; Caganova et al. 2013). EZH2 sustains the GC reaction in part by catalyzing the addition of the repressive H3K27me3 mark at genes that control proliferation (e.g., CDKN1A, CDKN1B) and plasma cell differentiation (e.g., IRF4, PRDM1) to create bivalent promoters that can then be rapidly reactivated when cells exit the GC (Béguelin et al. 2013). To study the biological consequences of the hotspot gain-of-function mutations Y641F and Y641N, two mouse models have been generated in which activation of the mutant allele can be directed to the GC following the excision of a lox-stop-lox cassette by the Cγ1-Cre recombinase: a conditional Ezh2^Y641F knock-in driven by the endogenous Ezh2 promoter (Béguelin et al. 2016) and a transgenic Ezh2^Y641N allele controlled by the ColA1 promoter (Béguelin et al. 2013). In both, expression of mutant Ezh2 caused massive GC hyperplasia after immunization, which was sustained by enhanced proliferation and blockade of terminal differentiation, and required the functional cooperation between EZH2 and the BCL6/BCOR repressor complex (Béguelin et al. 2013). This phenotype is also consistent with the detection of abundant H3K27me3 levels at the promoter of EZH2 target genes in the GC B cells from these animals, analogous to EZH2-mutant DLBCL cell lines and human tumors. Although Ezh2 mutant mice do not develop B-cell lymphomas, the adoptive transfer of VavP-Bcl2 bone marrow cells transduced with Ezh2^Y641F vectors were associated with accelerated lymphomagenesis (Béguelin et al. 2013; Ennishi et al. 2019a). Additionally, EZH2 mutation cooperates with deregulated BCL6 expression to yield a transplantable GC-derived DLBCL-like disease that can be used for syngeneic immune studies (Béguelin et al. 2016; Mondello et al. 2020). Of therapeutic interest, tumors developing in the above models display significantly lower cell surface levels of both MHC-I and MHC-II, which are accompanied by an immune-“cold” environment with reduced T-cell infiltrate and could be rescued by treatment with EZH2 inhibitors (Ennishi et al. 2019a).

Truncating mutations of TET2, the most common event in age-associated clonal hematopoiesis (Busque et al. 2012) and a genetic hallmark of myeloid neoplasms (Moran-Crusio et al. 2011), are also detected in 3%–10% of FL/tFL (Pasqualucci et al. 2014) and 6%–12% of DLBCL (Reddy et al. 2017; Chapuy et al. 2018; Schmitz et al. 2018). To study the in vivo role of these alterations in lymphomagenesis, murine strains were engineered to model the loss of Tet2 in HSCs or at later stages of B-cell development (Dominguez et al. 2018). Interestingly, Tet2 deficiency facilitated the expansion of GC B cells in Vav-Cre and CD19-Cre conditional KO mice, but not when Tet2 was deleted in the GC. These abnormal cells do not evolve into clonal DLBCL; however, effacement of the splenic architecture due to enlarged follicles or diffuse lymphoid infiltrates was observed when GC-specific Tet2 deletion was combined with BCL6 deregulation, even though the molecular features of these tumors, which are negative for several mature B-cell markers (CD23, CD21, IgM, IgD), require further characterization. This work also revealed a link between TET2 and CREBBP in orchestrating the GC exit transcriptional program through the activation of enhancer domains.

Collectively, the above studies illustrate a partially overlapping mechanism by which mutations in epigenetic modifiers can reprogram the epigenome of the precursor cancer cell at specific gene sets to enable malignant transformation, in cooperation with BCL2 deregulation, and establish a link between epigenetic dysregulation and immune escape that could be exploited for therapeutic intervention.

Modeling TNFRSF14 Loss

The GC microenvironment provides essential signals for the survival and differentiation of normal B cells and is postulated to play a particularly important role during FL development, as immune and stromal cells create a permissive niche to support the malignant B-cell population (Scott and Gascoyne 2014; Lamaison and Tarte 2019). The TNFRSF14 gene, encoding the HVEM receptor, is one of the most frequently mutated genes in FL and GCB-DLBCL. The contribution of its genetic loss to lymphomagenesis was studied by using an shRNA-knockdown strategy in the VavP-Bcl2 HPC adoptive transfer system (Boice et al. 2016). Although the developmental stage of the Hvem knockdown may not recapitulate the sequence of transforming events occurring in the human tumors, only a minority of CD4⁺ and CD8⁺ T cells were found to express the shHvem hairpin, whereas shHvem-expressing B-lymphoma cells were significantly enriched. Hvem knockdown caused marked B-cell expansion and accelerated the development of BCL2-driven FL. Mechanistically, this model was critical in demonstrating a dual role for HVEM loss during malignant transformation, which depends on its ability to (1) directly stimulate BCR signaling and B-cell proliferation, a cell-autonomous and BTLA-dependent effect exemplified by the activation of several molecules related to the BCR pathway such as SYK, BTK, BLNK, and their downstream targets ERK and IκB; and (2) induce a tumor-supportive microenvironment through the increased production of TNF-family cytokines that act as stroma-activating factors. The latter is consistent with the observation of an aberrant lymphoid stroma activation in FL that is more prominent in both murine and human TNFRSF14-deficient tumors. Moreover, this study offered new therapeutic perspectives for this group of tumors by showing that the administration of the soluble HVEM ectodomain protein could restore tumor suppression.

Modeling Missense RRAGC Mutations

The RRAGC gene encodes the RagC GTPase, an activator of a nutrient-sensing pathway that drives cellular anabolism (Efeyan et al. 2015). RagC forms a heterodimeric complex together with RagA that activates mTORC1 upon ample nutrient availability, leading to cell growth. To mimic the most common FL-associated RRAGC mutations in the mouse germline, the CRISPR–Cas9 genome engineering technology was employed. This study revealed that Rragc-mutant B cells show a partial insensitivity to nutrient withdrawal, which led to accelerated FL tumorigenesis on a VavP-Bcl2-transgenic background (Ortega-Molina et al. 2019). Mechanistically, RagC-mutant cells were found to exhibit suppression of apoptosis and a decreased requirement on microenvironmental signals provided by Tfh cells, which control the GC reaction. These data are consistent with a model in which RRAGC mutations increase GC B-cell fitness by inducing activation of the mTORC1 pathway; because of their competitive advantage over unmutated GC B cells, these mutant premalignant GC B cells continue to undergo cycles of selection and proliferation in the GC, which in turn could facilitate the acquisition of additional genetic alterations, ultimately transforming into a bone fide FL.

Modeling MEF2B-Activating Mutations

The MEF2B transcription factor is specifically expressed at high levels in GC B cells, where it controls the activity of a broad transcriptional network of relevance to the physiology of the GC reaction and also including the GC master regulator BCL6 (Ying et al. 2013). This function is perturbed in ∼15% of FL and DLBCL because of gain-of-function mutations that can be broadly categorized into two groups: (1) amino acid substitutions in the protein amino-terminal portion, encoding for the DNA-binding domain—these mutations prevent its physical interaction to components of the HUCA complex as well as to several class IIa HDACs, all of which normally serve as negative modulators of MEF2B activity; and (2) truncating mutations in the carboxy-terminal portion of the MEF2B protein, which are postulated to interfere with negative regulatory mechanisms of its activity mediated by post-translational modifications (e.g., sumoylation), although a comprehensive characterization of this second group of alterations is still lacking (Morin et al. 2011; Ying et al. 2013). Because MEF2B transcription is specifically induced in the initial stages of GC B-cell commitment, and thus slightly earlier than Cγ1-promoter activation, a conditional knock-in mouse model for the most common D83V missense mutation was engineered using the CD21-Cre deleter strain (Brescia et al. 2018). Compared to control littermates, Mef2b^+/D83V;CD21-Cre mice responded to polyclonal antigenic stimulation with a significant increase in GC B cells (a phenotype also observed upon Cγ1-Cre-mediated activation of the mutant allele), which evolved over time toward clonal FL or DLBCL in ∼20% of cases. This phenotype becomes fully penetrant when combined with BCL2 deregulation in the BCL2-Ig mice, thus establishing MEF2B as a fundamental player in the physiologic GC reaction and a driver oncogene in lymphomagenesis.

DIFFUSE LARGE B-CELL LYMPHOMA

The most common B-cell lymphoma in the Western world, DLBCL represents a heterogeneous disease comprising phenotypically and molecularly distinct entities that are associated with distinct clinical outcome and up to five genetic subsets defined based on co-occurring mutational events (Chapuy et al. 2018; Schmitz et al. 2018). Many of the genetic alterations that segregate with GCB-DLBCL, and particularly with its genetic subtype EZB (for EZH2-BCL2) or Cluster 3, are shared with FL. Indeed, as summarized in the previous section, GEMMs for these alterations develop a spectrum of lymphoproliferative disorders ranging from in situ FL to overt DLBCL. Additional DLBCL-associated genetic lesions that have been successfully and faithfully reconstructed in the mouse include those targeting BCL6, GNA13, and several genes implicated in the pathogenesis of ABC-DLBCL such as MYD88, PRDM1, and components of the NF-κB pathway.

Deregulation of BCL6 Expression

As the master regulator of the GC reaction, BCL6 is expressed at high levels in GC-derived lymphoid malignancies, including FL and GCB-DLBCL, and also represents a biological dependency in these tumors, which require its activity for their proliferation and survival (Basso and Dalla-Favera 2010; Valls et al. 2017). The first in vivo model recapitulating the genetics and therefore the biology of DLBCL was obtained by engineering a knock-in allele in which expression of an HA-tagged BCL6 cassette is driven by the endogenous immunoglobulin Iµ promoter, thus mimicking a common chromosomal translocation in human DLBCL and FL (Cattoretti et al. 2009). The broader activity of the juxtaposed Ig promoter/enhancer sequences, which extends beyond the GC stage, disrupts the normally restricted pattern of BCL6 expression, allowing the escape from multiple negative modulatory signals that normally induce its down-regulation at the GC exit (Cattoretti et al. 2009; Basso and Dalla-Favera 2010). The translocation may also provide marginal zone B cells, which normally lack BCL6 expression, with ectopic access to its proto-oncogenic activity. This scenario is supported by the observation that BCL6 translocations are particularly enriched in a subset of ABC-DLBCL that belongs to the BN2(BCL6-Notch2) or Cluster 1 genetic subgroup, for which a marginal zone B-cell origin has been postulated (Chapuy et al. 2018; Schmitz et al. 2018).

Independent of whether caused by chromosomal translocations or simply reflecting the origin from a GC B cell, the net effect of BCL6 continuous expression is the aberrant maintenance of multiple biological programs that are normally orchestrated by BCL6 to sustain the specialized function of the GC, including the negative regulation of anti-apoptotic and DNA damage responses, plasma cell differentiation, and a variety of receptor signaling pathways (Ci et al. 2009; Basso et al. 2010). Blocked in this environment, B cells continue to be exposed to potentially deleterious events, exemplified by the high proliferative rate, replication stress, and the DNA breaks associated with SHM and CSR. Consistently, IµHABCL6 mice display massive GC hyperplasia with an increased DZ:LZ ratio, even in the absence of antigenic stimulation (Cattoretti et al. 2005). The oncogenic link between BCL6-deregulated expression and lymphomagenesis was conclusively proven by the evidence that IµHABCL6 mice develop an array of lymphoproliferative diseases (LPDs) culminating in overt DLBCL in 40%–60% of cases and displaying evidence of AID-dependent aberrant SHM as well as Myc-IgH translocations (Cattoretti et al. 2005; Pasqualucci et al. 2008). Moreover, deregulated BCL6 expression led to a shift in the phenotype of λMYC-driven lymphomas reflecting their origin from GC-experienced cells and the requirement of AID (Pasqualucci et al. 2008).

Disruption of the Gα13 Signaling Pathway

Unlike most lymphocytes, GC B cells are strictly confined to the GC microenvironment. What prevents the exit of GC B cells into circulation is an inhibitory circuit that entails the engagement of G-protein-coupled receptors S1PR2 and P2RY8 by S1P ligands (Green and Cyster 2012). Inhibitory signals emanating from these receptors are transmitted via two G-proteins abundantly transcribed in GC B cells—namely, Gα12 and Gα13—to the effector molecule ARHGEF1, which ultimately acts by suppressing both AKT and cell migration (Green and Cyster 2012). In almost one-quarter of DLBCL, as well as in BL, the Gα13 pathway is disrupted by deleterious mutations affecting the genes GNA13 and more rarely S1P2R and ARHGEF1 (Muppidi et al. 2014). The GNA13-deficient state of GC B cells was modeled in the mouse by either crossing a Gna13 conditional knockout allele with an AID-Cre transgenic strain (Healy et al. 2016) or utilizing a mixed bone marrow chimera approach (Gna13^fl/fl;Mb1-Cre) (Muppidi et al. 2014). In both systems, Gα13-deficient mice manifested increased numbers of GC B cells in the context of a disordered GC architecture with altered DZ/LZ distribution, increased IgV SHM activity, and abnormal B-cell migration behavior. The persistence of GC B cells that display impaired caspase-mediated cell death led to an increased risk of lymphoma development, as documented by the insurgence in a subset of cases of massive mesenteric lymphoadenopathies of GC origin. Clonal B-cell lymphomas reproducing morphologic, phenotypic, and genetic characteristics of the human DLBCL were also detected in 50% of mice deficient for S1PR2, supporting the critical role of this pathway as a lymphoma tumor suppressor (Cattoretti et al. 2009). Interestingly, deficiency of Gα13 but not of S1PR2 led to systemic dissemination in the lymph and blood, suggesting the existence of additional G-protein-coupled receptors associated with the regulation of GC confinement and prompting the discovery of P2RY8, which is also mutated in some DLBCL (Muppidi et al. 2014). These studies were instrumental in elucidating the mechanism by which GNA13-deficient GC B cells can exit the GC niche to spread systemically and demonstrated a tumor-suppressor function for the Gα13 signaling pathway through its dual effect in the control of B-cell positioning and AKT activation.

Deletion of FBXO11

F-box protein 11 (FBXO11) is a member of the F-box protein family that functions as the substrate-recognition subunit of SKP1-cullin-1-F-box-protein (SCF) E3 ligase complexes, leading to ubiquitylation and degradation of target proteins (Skaar et al. 2013). Of direct relevance to lymphoma, the BCL6 master regulator and the BLIMP1 protein were both identified as FBXO11 substrates. Indeed, monoallelic mutations and/or deletions of the FBXO11 gene, detected in 6% of human DLBCLs, were shown to result in increased BCL6 levels, suggesting a contributing role during DLBCL pathogenesis (Duan et al. 2012). A conditional, GC-specific knockout mouse model was constructed by flanking the gene exon 4 with loxP sites, recombination of which results in a translational reading frameshift with production of a truncated FBXO11 protein (Schneider et al. 2016). Fbxo11^fl/fl;Cγ1-Cre mice display enlarged GCs with significantly increased BCL6 protein levels and reduced ability to enter the plasmablastic differentiation pathway, resembling the phenotype observed in mice with constitutive BCL6 expression or Prdm1 deletion. Aged, chronically immunized mice develop various B-cell lymphoproliferative phenotypes ranging from expanded lymphoid follicles through LPDs disrupting the lymphoid architecture to overt DLBCL, although at very low frequency. Such low tumor penetrance implies that additional alterations are required to drive full transformation, together with FBXO11 inactivation. Nonetheless, this model confirmed a tumor-suppressor role for FBXO11 in lymphomagenesis, which is linked at least in part to altered BCL6 protein stability, even though multiple proteins are likely to be recognized as substrates of FBXO11-mediated degradation, in addition to BCL6 and BLIMP1.

Biallelic Loss of PRDM1/BLIMP1

A distinctive feature of ABC-DLBCL is the presence of structural alterations that lead to genetic or epigenetic inactivation of the master plasma cell regulator BLIMP1 (also known as PRDM1), with 20% of cases carrying biallelic disruptive mutations and/or focal deletions of the BLIMP1 locus, and an additional subset of tumors showing mutually exclusive transcriptional silencing by deregulated BCL6 (Pasqualucci et al. 2006; Tam et al. 2006). When engineered in the mouse, conditional B-cell-specific deletion of Blimp1 (Blimp1^fl/fl;CD19-Cre and Blimp1^fl/fl;Cγ1-Cre) induced a block in plasma cell differentiation and the development of DLBCLs recapitulating the molecular footprint of the human ABC-DLBCL (Mandelbaum et al. 2010). As in other lymphoma models, the long latency and the clonality of DLBCLs in Blimp1 conditional KO animals indicate that oncogenic events affecting other pathways collaborate with inactivation of this transcription factor to foster a full neoplastic phenotype. One important contributor to this process is the NF-κB transcriptional complex, which is constitutively active in virtually all ABC-DLBCLs and is targeted at multiple levels by genetic alterations in more than one-half of these cases (Compagno et al. 2009), frequently together with BLIMP1 mutations (Shaffer et al. 2012). Accordingly, the DLBCLs that develop in Blimp1 conditional KO mice display nuclear active NF-κB (Mandelbaum et al. 2010); moreover, a mouse model with combined disruption of Blimp1 and enforced canonical NF-κB activation, obtained via a constitutively active IKK2 protein (R26Stop^FLIkk2ca;Cγ1-Cre) that can be induced in GC B cells, showed enhanced GC responses and succumbed to clonal lymphoid tumors reminiscent of human ABC-DLBCL (Calado et al. 2010).

Constitutive Activation of the NF-κB Signaling Pathway

During normal GC responses, the canonical (RELA/p50 and c-REL/p50) and noncanonical (RELB/p52) NF-κB signaling pathways have been shown to play distinct roles (Heise et al. 2014; De Silva et al. 2016), which are hijacked by lymphoma cells for their own growth advantage. In most ABC-DLBCL cases, the activity of the canonical NF-κB transcription complex is sustained by the presence of genetic alterations affecting multiple genes that encode for positive or negative regulators of the BCR, CD40 receptor, and TLR signaling cascades (Lenz et al. 2008; Compagno et al. 2009; Davis et al. 2010; Ngo et al. 2011). Among these, the TLR adaptor protein MYD88 is mutated in >30% of patient samples (Ngo et al. 2011). Consistently, AID-Cre-, CD21-Cre-, or CD19-Cre-driven expression of a Myd88^L252P allele corresponding to the most common human activating mutation (L265P) promotes the occurrence of indolent LPDs and occasional tumors that share several traits with human ABC-DLBCL (Knittel et al. 2016). Of note, a synergistic cross talk was observed between this mutation and CD79B mutations in a murine adoptive transfer experiment, exemplified by the accumulation of autoreactive cells (Wang et al. 2017); this finding is interesting in view of the frequent co-occurrence of CD79B mutations and MYD88-L265P alleles in ABC-DLBCL cases and the reported role for self-antigens in the survival of ABC-DLBCL cells via chronic activation of their BCR-signaling pathway (Young et al. 2015).

In a smaller subset of DLBCL, the alternative (noncanonical) NF-κB signaling cascade is also activated, as documented by the nuclear translocation of p52 (Compagno et al. 2009). This can be explained in part by the presence of truncating mutations/deletions of the TRAF3 gene, often coexisting with BCL6 translocations. TRAF3 encodes for a negative regulator of the pathway, involved in the degradation of the NF-κB inducing kinase NIK. Accordingly, enforced expression of NIK and BCL6 in the GC, as obtained by conditional mutagenesis in the IμHABcl6;Nik^stopFL;Cγ1-Cre mouse model, caused GC hyperplasia with blockade of terminal differentiation and ultimately premature death in 100% of the animals, because of IRF4-positive DLBCL (Zhang et al. 2015a). Notably, Nik^stopFL;Cγ1-Cre mice display overt plasma cell hyperplasia but do not succumb to tumors; thus, the oncogenic function of the alternative NF-κB pathway may require the concomitant disruption of terminal B-cell differentiation, which in this case was achieved by deregulated BCL6 expression, and thus appears to be analogous to the synergistic phenotype observed in the compound Blimp1^fl/fl;R26Stop^FLIkk2ca;Cγ1-Cre model (Calado et al. 2010).

CHRONIC LYMPHOCYTIC LYMPHOMA/SMALL LYMPHOCYTIC LYMPHOMA

CLL/SLL is an indolent B-cell malignancy with the tumor cells characteristically expressing the CD5 and CD23 antigens (Kipps et al. 2017). Compared to other non-Hodgkin lymphomas, CLL shows a different spectrum of genetic alterations, which mostly comprise chromosomal deletions (13q14, ATM, and TP53) or amplifications (trisomy of chromosome 12). Next-generation sequencing analyses could identify additional recurrent mutations in CLL/SLL, most notably those that target the NOTCH1, MYD88, and the SF3B1 genes (Fabbri and Dalla-Favera 2016). Additionally, genome-wide association studies identified a single-nucleotide polymorphism (SNP) in a noncoding region of IRF4, a major transcription factor involved in B-cell development (De Silva et al. 2012), as a susceptibility locus for CLL (Di Bernardo et al. 2008).

To date, several mouse models of CLL have been obtained by reconstructing the above mentioned somatic genetic aberrations or germline SNPs. Additionally, numerous strains exist that develop CLL-like diseases but have been produced by introducing genetic changes that are not observed in the human tumors or are activated at different developmental stages compared to the putative normal counterpart of CLL. Although we refer the reader to other reviews for a detailed description of these models (Pekarsky et al. 2007; Chen and Chiorazzi 2014; Simonetti et al. 2014), we mention here the Eµ-TCL1 transgenic mice (Bichi et al. 2002) as they are widely used in the field to study the role of the microenvironment in the disease process as well as for preclinical studies, either alone or in combination with other alterations (Johnson et al. 2006; Simonetti et al. 2014). The Eµ-TCL1 model, however, does not recapitulate the genetics and, thus, possibly the biology of this common cancer. This circumstance needs to be considered when interpreting the findings for the human disease, although the resemblance of the lymphoproliferations developing in these mice to the more aggressive form of CLL makes them an amenable model for IgV unmutated CLL with unfavorable outcome.

Deletion of chromosomal region 13q14 occurs in >50% of CLL cases and at a lower frequency also in other lymphoma subtypes (Liu et al. 1995). The 13q14 region is highly conserved in the mouse genome at chromosomal region 14qC3 and has been further mapped into two characteristically deleted regions. The so-called minimal deleted region (MDR) encompasses two microRNAs—namely, mir-15a/16-1—located within an intron of the deleted in leukemia 2 (DLEU2) gene that encodes a long noncoding RNA; the mouse 13q14 region also includes the protein-coding Dleu5 gene. The commonly deleted region (CDR), present in a sizable number of CLL/SLL cases, is larger and, besides the MDR, includes the protein-coding DLEU7 and RNASEH2B genes in both humans and mice. Three transgenic mouse lines have been generated in which only mir-15a/16-1, the MDR, or the CDR were deleted in B cells (Klein et al. 2010; Lia et al. 2012). Deletion of the microRNAs alone caused the development of a CLL-like disease at low penetrance (∼25%), demonstrating the tumor-suppressor role of mir-15a/16-1 in vivo, as proposed in part following the observation that (i) germline mutations interfering with their normal expression occurred in some CLL patients (Calin et al. 2002, 2005), and (ii) an NZB mouse strain with a germline mutation in the 3′ flanking region of pre-miR-16-1 results in decreased expression of the mature microRNA (Raveche et al. 2007). The additional deletions of Dleu2 and Dleu5 in MDR mice increased the disease penetrance and exacerbated the disease course. Of note, besides CD5⁺ lymphoproliferations, also CD5⁻ lymphomas were observed. CDR-deleted mice showed a different spectrum of lymphoproliferations compared with the other two lines, as they predominantly developed CLL-like lymphoid neoplasms and a more aggressive disease course. It is not clear whether the target cell of malignant transformation is the same for CLL developing in mice and humans. One clear difference is that all CD5⁺ lymphoproliferations developing in the transgenic lines express unmutated IgV region genes, whereas, in humans, IgV mutated cases comprise a considerable fraction of all CLL diagnoses. However, all three 14qC3 mouse models expressed stereotypic antigen receptors, a characteristic feature of human CLL (Kipps et al. 2017), which indicates a critical role of antigen in the expansion of the CLL clone. One may therefore conclude that, at least with regard to the IgV unmutated cases, a B-cell subset with similar biological functions is targeted in both humans and mice.

The SNP identified in human CLL has been associated with a down-regulation of IRF4 expression, suggesting an atypical tumor-suppressor role for IRF4 in CLL development (Di Bernardo et al. 2008). This hypothesis has been investigated in a mouse model in which an antibody expressing a rearranged IgV region gene of the Vh11 family that is frequently expressed in CD5⁺ B cells was knocked into the Ig locus and crossed onto an IRF4-deficient background. Irf4^−/−Vh11 knock-in mice developed a CLL-like disease with complete penetrance (Shukla et al. 2013); a follow-up study reported elevated NOTCH2 expression and hyperactivation of NOTCH signaling as the underlying mechanism (Shukla et al. 2016). Because Irf4^−/− mice have multiple immune deficiencies as a result of the critical roles of IRFs in a number of immune cell types, the determination of the extent to which IRF4 deficiency affects CLL development in a B-cell-intrinsic fashion would require B-cell-specific deletion of Irf4 on the Vh11 background. Such a model would allow us to investigate the proposed role of reduced IRF4 expression in promoting CLL development by altering the migration properties of B cells (Simonetti et al. 2013), as a direct consequence of elevated NOTCH2 activity (Shukla et al. 2016). A B-cell-intrinsic role of IRF4-deficiency in CLL development has recently been demonstrated via conditional Irf4 knockout on the Eμ-TCL1 background (Asslaber et al. 2019).

Mutations in the splicing factor SF3B1 are found in 10% of CLL cases (Wang et al. 2011). To determine the functional consequences of the most commonly occurring SF3B1 point mutation, K700E, a conditional Sf3b1-K700E knock-in allele, was created that allowed to express the mutated allele in a B-cell-specific fashion using CD19-Cre mice (Yin et al. 2019). Expression of the mutated SF3B1 protein led to an enlargement of the marginal zone, although the total number of B cells was reduced, possibly in relationship to the induction of cellular senescence. Indeed, these animals did not develop B-lymphoproliferations unless bred on a conditional ataxia telangiectasia mutated (ATM)-deficient background, to reproduce the frequent co-occurrence of SF3B1 mutations with deletion of chromosomal region 11q, encompassing the gene, in CLL (Wang et al. 2011; Yin et al. 2019). Atm deletion could overcome the SF3B1-K700E-induced cellular senescence and led to the development of a clonal CLL-like disease in about one-half of the animals by 24 mo of age. Phenotypically, the tumor cells resembled human CLL cells with regard to genome instability and the dysregulation of CLL-associated cellular processes. This included down-regulation of BCR signaling, which is also observed in human CLL cells with SF3B1 mutations. Consistent with these observations, CLL cells were more susceptible to the BTK inhibitor ibrutinib, which acts downstream of the BCR. Together, this model demonstrates the practicability of modeling specific tumor-associated genetic alterations to gain insights into the disease-underlying pathogenic mechanisms and to uncover potential Achilles’ heels for therapy.

CONCLUSIONS

The advantages and versatility offered by GEMMs have revolutionized the study of cancer biology and will continue to be an invaluable resource to obtain critical insights into the multifaceted pathogenic mechanisms that underlie B-cell malignancies. However, no single model is likely to reproduce the complexity of human cancer. Moreover, the accelerated discovery of new cancer genes and the multitude of genetic interactions that are implicated in the malignant transformation process, as emerged from large-scale genome sequencing studies, pose a significant challenge to the capacity of traditional GEMMs for providing rapid advances. Finally, the plethora of novel therapeutic agents and concepts that are being considered for preclinical testing, alone or in combination, warrants the need for more effective, high-throughput approaches to modeling cancer in a faithful manner. Tumor organoids, PDXs, and the advent of increasingly sophisticated technologies such as the CRISPR–Cas9 gene editing technique may help overcome some of these limits—for example, by improving the speed at which multiple genetic changes can be simultaneously studied and by reproducing the complex interactions between the microenvironment of the patient and the malignant cells. Further efforts will also be needed by the mouse-modeling community to reduce the costs of these studies and provide more uniform systems for translational interrogation. Understanding the strength and limitations of each model remains instrumental to maximally leverage these resources.