The mechanisms of human lymphoid chromosomal translocations and their medical relevance

Abstract

The most common human lymphoid chromosomal translocations involve concurrent failures of the recombination activating gene (RAG) complex and Activation-Induced Deaminase (AID). These are two enzymes that are normally expressed for purposes of the two site-specific DNA recombination processes: V(D)J recombination and class switch recombination (CSR). First, though it is rare, a low level of expression of AID can introduce long-lived T:G mismatch lesions at 20–600 bp fragile zones. Second, the V(D)J recombination process can occasionally fail to rejoin coding ends, and this failure may permit an opportunity for Artemis:DNA-dependent kinase catalytic subunit (DNA-PKcs) to convert the T:G mismatch sites at the fragile zones into double-strand breaks. The 20–600 bp fragile zones must be, at least transiently, in a single-stranded DNA (ssDNA) state for the first step to occur, because AID only acts on ssDNA. Here we discuss the key DNA sequence features that lead to AID action at a fragile zone, which are (a) the proximity and density of strings of cytosine nucleotides (C-strings) that cause a B/A-intermediate DNA conformation; (b) overlapping AID hotspots that contain a methyl CpG (WRCG), which AID converts to a long-lived T:G mismatch; and (c) transcription, which, though not essential, favors increased ssDNA in the fragile zone. We also summarize chromosomal features of the focal fragile zones in lymphoid malignancies and discuss the clinical relevance of understanding the translocation mechanisms. Many of the key principles covered here are also relevant to chromosomal translocations in non-lymphoid somatic cells as well.

Introduction and overview of chromosomal translocations

A chromosomal translocation is the most common first essential step for most human lymphoid malignancies (lymphoid leukemias and lymphomas). It usually involves two phases: the breakage phase (at each of two chromosomes) and the rejoining phase (Figure 1(A)) (Tsai et al. 2008; Lieber 2016). Non-homologous end joining (NHEJ) is active throughout the entire cell cycle and is the major pathway employed by somatic cells to repair DNA double-strand breaks (DSBs) in the rejoining phase (Chang et al. 2017; Zhao, Rothenberg, et al. 2020). Normally, NHEJ joins the correct ends from each double-strand break (DSB) in nearly all cases. Under rare abnormal circumstances, NHEJ can also join the incorrectly broken chromosome ends from two DSBs that arise concurrently, and this can result in a chromosomal translocation (Figure 1(B)).

Figure 1.

Illustration of factors causing double-strand DNA breaks (DSBs) and chromosomal translocation consequences. (A) Causes and repair of DSBs. Physiological and pathological causes of DSBs in mammalian somatic cells are listed at the top. At any time in the cell cycle, DSBs can be repaired by non-homologous DNA end-joining (NHEJ). During S and G2 of the cell cycle, homology-directed repair is common because the two sister chromatids are in close proximity, providing a nearby homology donor. Homology-directed repair includes homologous recombination (HR) and single-strand annealing (SSA). Proteins involved in the repair pathways are also shown. RAG: recombination activating gene; AID: Activation-Induced Deaminase; UDG: uracil-DNA glycosylase; APE1: apurinic-apyrimidinic endonuclease 1; DNA-PKcs: DNA-dependent kinase catalytic subunit; Pol: DNA polymerase; XLF: XRCC4-like factor; NBS1: Nijmegen breakage syndrome 1; MRE11A: meiotic recombination 11 homolog A. (B) Consequences of chromosomal translocations. The reciprocal chromosomal translocation results in two derivative chromosomes with exchanged segments. In the case where each chromosome contains one centromere as shown in the left panel, the two derivative chromosomes are usually stable. However, if one of the two derivative chromosomes contains two centromeres and the other has none as shown in the right panel, they are usually lost or unstable during cell growth.

The incorrect exchange of the DNA ends between the two DSBs could occur due to the broken ends of one DSB diffusing apart; then each DNA end from the first DSB may be joined by random association to the broken DNA ends at a second DSB on a different chromosome. After joining all four ends, the two resulting translocated chromosomes are called derivative chromosomes. In cases where both derivative chromosomes have one centromere, these chromosomes are usually stable (Figure 1(B), left side). In cases where one chromosome has two centromeres and the other has none, then the latter chromosome is often lost from progeny cells after cell division, due to the failure of the chromosome to align at the metaphase plate (Figure 1(B), right side). But any derivative chromosome bearing two centromeres will undergo breakage-fusion-bridge cycles, resulting in further variation of that derivative chromosome during subsequent cell divisions (Murnane 2006).

The DNA breakage phase has been the most difficult to study and understand, and that is the primary focus of this review. DNA breakage inside somatic cells can be due to physiologic or pathologic causes (Figure 1(A)) (Lieber 2010; Lieber et al. 2010). V(D)J recombination (Figure S1) and class switch recombination (CSR) (Figure S2) are two major physiologic processes that involve DNA breakage in vertebrates and occur only in lymphoid cells. The recombination activating gene (RAG) complex (a heterotetramer composed of two RAG1 and two RAG2 molecules) is the nuclease that initiates DNA breakage in V(D)J recombination at the early stages of B or T cells to generate a diverse repertoire of immunoglobulins and T cell receptors (Schatz et al. 1989; Oettinger et al. 1990). Activation-Induced Deaminase (AID) is involved in CSR (Figure S2) and somatic hypermutation (SHM) (Figure S3) within germinal center B cells to generate diversity in the variable domain exons and to vary the constant region of the antibody heavy chain and generate diversity in the variable domain exons (Muramatsu et al. 2000). Most of the chromosomal translocations in human lymphoid malignancies are due to the incorrect rejoining of a DNA break initiated by the RAG complex or AID at the immunoglobulin heavy (IGH) locus and a second break at a non-IGH locus, typically an oncogene (Figure 2) (Lieber 2016). The breakage initiated by the RAG complex or AID is located at specific DNA motifs due to the sequence preferences of each of these lymphoid-specific enzymes, and these will be discussed later.

Figure 2.

Common chromosomal translocations in B cells. Most of the lymphoid translocations in the early B cell (pro-B/pre-B) stage (A,B,D,E) involve a double-strand DNA break (DSB) that is a RAG-induced event (termed RAG-type break) at the immunoglobulin locus and a second DSB that is an AID-induced event (termed AID-type break) on an oncogene. The E2A-PBX1 translocation in early B cells (C) involves an AID-type break on E2A and a second break due to random causes, such as oxidative stress, ionizing radiation, or others on PBX1. Mature B cell translocations (F,G) commonly involve an AID-type event on one chromosome and a second event involving a failed IGH class switch recombination (CSR) event, which is a physiologic AID-type event. MALT1: mucosa-associated lymphoid tissue lymphoma translocation 1; CRLF2: cytokine receptor-like factor 2.

Other than the physiologic somatic cell breakage enzymes in lymphoid cells, pathologic causes, such as oxidative free radicals, ionizing radiation (IR), or failed topoisomerase II reactions, are responsible for the DSB in translocations in all living cells. The breakage initiated by these general pathologic causes is quite often distributed over large regions because there is little or no sequence specificity. For example, IR damage can occur at any position on any chromosome. Growth advantages raise a select set of human translocations to clinical attention. For the growth advantage, the translocation may be anywhere within large regions [e.g. kilobases (kb) within a key intron or hundreds of kb in cases that result in increased expression via proximity to transcriptional or developmental regulatory elements]. In some cases of growth advantage, the fusion of two open reading frames encodes an oncogenic protein, often causing a survival or proliferation advantage, and thus a neoplasm. Examples of this are BCR-ABL (the Philadelphia chromosomal translocation) in myeloid malignancies and MLL translocations in some hematopoietic malignancies (Table 1).

Table 1.

Fragile zones of common human hematopoietic translocations.

20–600 bp fragile zones	>2 kb fragile zones	Large zones
BCL2 MBR	BCL6	BCR–ABL1
BCL2 ICR	MYC	MLL
BCL2 MCR		ETV6–RUNX1 (TEL–AML1)
BCL1 MTC		PBX1
23 bp E2A fragile zone		HLF
89 bp MALT1 fragile zone
311 bp CRLF2 fragile zone

The focus of this Review is on translocations in the first column, which represent hotspots of breakage in narrow regions of 20–600 bp. Translocations related to convergent transcription or the formation of R-loops are responsible for the ~2 kb fragile zones in BCL6 and MYC, respectively (Lu et al. 2015; Yu and Lieber 2019). Some translocations arise because they generate a fusion protein, and the break can be anywhere (or in large segments) of specific introns (as shown in the third column). CRLF2: cytokine receptor-like factor 2; ETV6: ETS variant 6; ICR: intermediate cluster region; MALT1: mucosa-associated lymphoid tissue lymphoma translocation 1; MBR: major breakpoint region; MCR: minor cluster region; MLL: mixed-lineage leukemia; MTC: major translocation cluster; RUNX1: runt-related transcription factor 1.

In the more common patient neoplasms, the DNA breakage mechanism at the non-IGH loci (the oncogenes) is extremely interesting because its highly focused nature suggests a consistent mechanism. Larger break regions of up to a few kb, such as BCL6 and MYC, are hotspots of DNA breakage in mature B cell translocations (Figure S5) (Tsai et al. 2008; Lu et al. 2013; Lu et al. 2015; Lieber 2016). Convergent transcription or R-loop formation has been reported to cause the breakage of BCL6 and MYC, respectively (Lu et al. 2015; Yu and Lieber 2019). In contrast to these, in early B cell translocations, a majority of these breaks, including the DSBs at BCL2, E2A (TCF3), BCL1 (CCND1), cytokine receptor-like factor 2 (CRLF2), and mucosal-associated lymphoid tissue lymphoma translocation 1 (MALT1), take place at AID hotspot motifs in narrow zones of 20–600 bp (Table 1; Figure 3; Figure S4), which are the focus of this current review (Tsai et al. 2008; Greisman et al. 2012; Cui et al. 2013; Lieber 2016; Pannunzio and Lieber 2017). While AID is most abundant in mature B cells, many studies have shown that there is a low level of AID expression in the pro- and pre-B cell stages (Mao et al. 2004; Han et al. 2007; Ueda et al. 2007; Kuraoka et al. 2009; Kuraoka et al. 2011; Kumar et al. 2013; Kelsoe 2014; Umiker et al. 2014; Cantaert et al. 2015). The low level of AID in the early stages of B cell development is sufficient to catalyze the rare but critical translocation events. AID can initiate a DNA lesion by deaminating cytosines in regions of transient or stable single-stranded DNA (ssDNA) (Bransteitter et al. 2003; Pham et al. 2003), resulting in an easily recognized and repaired U:G mismatch most of the time (Schmutte et al. 1995; Walsh and Xu 2006). Importantly, a T:G mismatch is generated by AID in the presence of methylcytosine at the CpG site. AID conversion of methylcytosine is about 10-fold less efficient than its activity on unmethylated cytosine (Bransteitter et al. 2003). However, compared with U:G mismatch which is repaired by the highly abundant uracil DNA glycosylase in cells, the T:G mismatch is more persistent since the thymine DNA glycosylase is in low abundance in mammalian cells and much less efficient than UDG in DNA damage processing (Schmutte et al. 1995; Walsh and Xu 2006). This long-lived T:G lesion can be converted to a DSB by the RAG complex or by the structure-specific nuclease, Artemis, in the context of the activated Artemis:DNA-dependent kinase catalytic subunit (DNA-PKcs) complex (Figure S6) (Ma et al. 2002; Tsai et al. 2008; Cui et al. 2013).

Figure 3.

Breakpoint distributions on oncogenes involved in lymphoid translocations in early B cells. (A) BCL2 breakpoints in BCL2-IGH translocations. The BCL2 breakpoints are scattered within a 30 kb region with three cluster regions (red starbursts). The breakpoints that do not fall into cluster regions are plotted as dark gray vertical lines. The 175 bp BCL2 major break region (MBR), located at the 3′ untranslated region (UTR) of BCL2, contains 50% of all BCL2 patient breakpoints. The breakage frequency within this 175 bp region is 300-fold higher than what one would expect to occur randomly. Patient breaks within the MBR are not uniformly distributed across the entire 175 bp, but rather are focused in three peaks centered around the CG motif. The 105 bp intermediate cluster region (ICR) and 561 bp minor cluster region (MCR) of BCL2 contains 13 and 5% of the BCL2 breakpoints, respectively. The breakpoints within MBR, ICR, and MCR are illustrated below the figure in zoomed-in views. Human lymphomas are clinically indistinguishable regardless of the position of the breakpoint within the 30 kb. (B) BCL1 breakpoints in BCL1-IGH translocations. The major translocation cluster (MTC) of BCL1 (also known as CCND1) is located 109 kb upstream of the CCND1 gene on chromosome 11 and contains 30% of all BCL1 breakpoints. The rest of the BCL1 breakpoints are scattered within a 344 kb intergenic region between the CCND1 gene at the telomeric end and the MYEOV gene on the centromeric side. (C) E2A breakpoints in E2A-PBX1 and E2A-HLF translocations. The E2A breakpoints in the E2A-PBX1 translocation (top panel) and E2A-HLF translocation (bottom panel) are illustrated. Over 75% of the E2A breakpoints are in a 23 bp region within E2A intron 16, making the 23 bp more than 400-fold more fragile compared with other regions of the same E2A intron. (D) MALT1 breakpoints in IGH-MALT1 and API2-MALT1 translocations. The breakpoints of MALT1 are focused on an 86 bp region upstream of the MALT1 gene in IGH-MALT1 translocations which are shown in the zoomed-in view. In contrast, the breakpoints of MALT1 in the API2-MALT1 translocation are scattered in a 29 kb region in several introns and exons of the MALT1 gene. For all figures, each triangle represents a single breakpoint from a patient. Triangles above and below the sequence are sequenced from two derivative chromosomes that resulted from the translocation. The CpG sites are highlighted with a red background. Within each fragile zone, the actual translocations occur at DNA sequence motifs, consisting of either the sequence CG (CpG) or mCG (when methylated) or WGCW where W = A or T. We are trying to determine why these small fragile zones are highly preferred for DNA breakage and translocation.

The central question remains why the AID activity is limited to such focal 20–600 bp regions on these oncogenes. We investigated the DNA and chromatin features around the fragile zones in E2A, BCL2, BCL1, MALT1, and CRLF2 to explore possible factors or sets of parameters that may contribute to the clustered DNA breakage, and that is a primary focus of this review. However, we also discuss other aspects of the translocation process, including a brief discussion of the joining phase, which we discuss first because it is the most clearly understood aspect and because some of the enzymes involved at the joining phase also contribute during the more complicated DNA breakage phase of the translocation process.

The joining phase of chromosomal translocations

After any single DSB occurs, NHEJ is the predominant pathway available through the whole cell cycle for repairing the DNA damage (Beucher et al. 2009; Ghezraoui et al. 2014). The two resulting DNA ends from a DSB are repaired, usually with nucleotide loss and modification at the joining site before restoration of the overall configuration of the chromosome. For chromosomal translocations, two DSBs are involved and result in a pathologic configuration of the chromosomal segments. The enzymes in the NHEJ pathway are the predominant ones involved in making the broken DNA ends chemically suitable for ligation, regardless of whether the correct or the translocated chromosomal configuration results.

In NHEJ, a specific nuclease, any of several polymerases, and a ligase participate in the rejoining phase (Chang et al. 2017). Direct ligation of the two ends is usually blocked by end incompatibility and chemical modification. Ku binding at the DNA ends after breakage prevents excessive DNA end resection (≥20 nt), which distinguishes NHEJ from other repair pathways (Chang et al. 2017; Pannunzio et al. 2018). The Artemis:DNA-PKcs complex, recruited to the Ku bound DNA ends, resects the DNA ends until short microhomology (≤4 nt) between the strands is exposed to facilitate proper ligation (Chang et al. 2017; Pannunzio et al. 2018).

Random nucleotide addition by polymerases is another mechanism that can generate microhomology. Members of the polymerase X family, including terminal deoxynucleotidyl transferase (TdT), polymerase mu (pol μ), and polymerase lambda (pol λ), are the main polymerases that generate ligatible ends. The inserted DNA sequence at the break junction provides some clues as to which polymerase added the nucleotides at the junction (Zhao, Rothenberg, et al. 2020). TdT functions in a template-independent manner with a preference for incorporating dCTP and dGTP (Gauss and Lieber 1996). TdT is only expressed in early B and T lymphocytes (Li et al. 1993), and one of its main activities is the random addition of nucleotides to ssDNA during V(D)J recombination to increase the antigen receptor and immunoglobulin diversity (Bertocci et al. 2006). The randomly added nucleotides by TdT during the rejoining phase are called N-nucleotides (N-nts) or N-regions (Landau et al. 1987). Pol μ and pol λ are involved in NHEJ in all cells, and they can incorporate nucleotides in both a template-independent and template-dependent manner (Wood et al. 2001; Maga and Hübscher 2003; Blanca et al. 2004; Ramadan et al. 2004; Tippin et al. 2004). The template-independent activity of pol λ is weaker compared with that of pol μ and TdT (McElhinny et al. 2005). When these nucleotides are incorporated in a template-dependent manner (by copying nts in either of the two DNA ends that are being joined) then nucleotide additions are often called T-nucleotides (T-nts) (see Supplemental Text for further discussion) (Lieber 2016; Zhao, Rothenberg, et al. 2020).

DNA ligase IV (Lig4) functions exclusively in NHEJ (Lieber 1999; Pannunzio et al. 2017). Two major steps are involved in the ligation of two broken dsDNA ends: the physical juxtaposition of DNA ends (synapsis) and the covalent ligation. The X-ray repair cross-complementing 4 (XRCC4) can stimulate the activity of Lig4 (Grawunder et al. 1997). The Lig4:XRCC4 complex (X4L4) is the central component for DNA ends synapsis and ligation (Grawunder et al. 1997; Zhao et al. 2019). When DNA ends contain at least 1 nt microhomology, pol μ can also mediate efficient synapsis for ligation (Zhao, Watanabe, et al. 2020). Once DNA ends are in proximity with proper configurations, the covalent ligation can occur very quickly (i.e. within seconds in purified biochemical systems).

For NHEJ, there are mechanisms involving Ku, XRCC4, and Lig4 that support end-to-end synapsis, and this is further stabilized by XRCC4-like factor (XLF) and paralog of XRCC4 and XLF (PAXX) (Zhao et al. 2019; Zhao, Rothenberg, et al. 2020). Filament formation by XRCC4 and XLF is a second mechanism that appears to support the proximity of two DNA ends before NHEJ (Hammel et al. 2010; Ropars et al. 2011; Reid et al. 2015; Brouwer et al. 2016; Lescale et al. 2016; Nemoz et al. 2018). A third mechanism for synapsis may involve DNA-PKcs. Crude extract studies provided indirect support for such a role, but conclusions from studies using purified systems have provided evidence both for and against such a role (Reid et al. 2015; Graham et al. 2016, 2018; Conlin et al. 2017; Reid et al. 2017; Wang et al. 2018; Zhao et al. 2019; Stinson et al. 2020; Zhao, Watanabe, et al. 2020; Chaplin, Hardwick, Liang, et al. 2021; Chaplin, Hardwick, Stavridi, et al. 2021; Chen et al. 2021).

Of note, the regional nuclear proximity of the two chromosomes may affect the rejoining phase but there is no evidence showing that chromosome proximity has any role in the DNA breakage phase in lymphoid neoplasms. It is not likely that the distance of two chromosomes inside the cells would affect the breakage of the DNA (Sunder and Wilson 2019; Wilson and Sunder 2020). If proximity is important, we would expect dominant translocation events between nearby chromosomes rather than most of the random translocations across the genome currently known. E2A-PBX1 translocations occur between chromosome 1 and chromosome 19 which are far away from each other according to the 3 D map of human nuclei (Bolzer et al. 2005).

Overview of causes of chromosome breaks

The DSBs in lymphoid cells can be induced by the general causes that apply to all living cells (Figure 1(A)). But more frequently in lymphoid cells, the lymphoid specific factors are involved, specifically the aberrant action of the RAG complex (during V(D)J recombination) or AID (during CSR at the IGH locus or during SHM at any of the immunoglobulin gene loci) (Mahowald et al. 2008; Tsai et al. 2008; Lieber 2010; Nussenzweig and Nussenzweig 2010; Gostissa et al. 2011). The two DSBs in lymphoid chromosomal translocations may have independent causes or the same cause. In one type, the RAG complex causes the breaks on both chromosomes, and these are referred to as RAG-type events at both chromosomal locations (that is, RAG-type/RAG-type translocations). In the most common cases, AID can initiate a lesion, causing a DSB on one chromosome (an AID-type break), and the RAG complex would cause the break at the other chromosome (RAG-type/AID-type translocation). In the third category of cases, both breaks can be AID-type. In a fourth category, the “uncertain-type” breaks, a break is neither RAG-type nor AID-type but generated by one of the general causes mentioned earlier (e.g. ionizing radiation, failed topoisomerase II reactions, oxidative damage; see Figure 1(A)). In all cases, if the two DSBs are on the same chromosome, then the same enzymes can cause a deletion within that chromosome, rather than a translocation.

The “uncertain-type” is predominantly the random breakage frequency for any phosphodiester bond. This frequency is not known with any precision and can only be commented in relative terms as follows. By comparison to the BCL2 and BCL1 translocations, if the frequency of random breakage was even 0.01 of the frequency of breakage at the BCL2 and BCL1 fragile zones, then these fragile zones would no longer be discernible. This is because the random breaks within and around the fragile zone would out-number the breaks with a defined cause (i.e. AID) within the fragile zone, and the boundaries of the fragile zone would become so diffuse that a fragile zone would no longer be definable.

RAG-type breaks at CAC motifs

The RAG complex normally cuts at the CAC of the heptamer since this is the invariant portion of the recombination signal sequence (RSS). RAG-type breaks can occur at CAC sequences at loci other than antigen receptors. RAG-type/RAG-type interstitial (intrachromosomal) deletion is one of the common outcomes that can lead to human T cell malignancies. Most notably, in pre-T cell acute lymphoblastic leukemia/lymphoma (T-ALL), the RAG complex breaks at both the SCL (also called TAL1) and SCL-interrupting locus (SIL; also known as STIL) locations are observed (Aplan et al. 1990; Lieber 1993). A small percentage of human B cell malignancies also appear to involve RAG-type/RAG-type interstitial deletions, such as the CRLF2-purinergic receptor P2Y8 (P2RY8) translocation on the X chromosome in some human pre-B ALLs (Weigert and Weinstock 2012). RAG-type/RAG-type translocations in human T-ALL, such as translocations between T cell receptor (TCR) locus and the SCL, LIM domain only 2 (LMO2), homeobox 11 (HOX11; also known as TLX1), and TTG1 (also known as LMO1) loci, are less common than the interstitial deletions (Raghavan et al. 2001; Norris and Stone 2008). The IGH-CRLF2 translocation is an example of a RAG-type event at IGH and an AID-type event at CRLF2 in B cell ALL (Tsai, Yoda, et al. 2010).

In summary, RAG-type/RAG-type translocations account for a majority if not all consistent translocations in T-ALLs; notably, T cells have no AID expression. In contrast, the co-expression of the RAG complex and AID is responsible for the most common translocations in B cells, and this is a very critical point that cannot be overemphasized. Given the involvement of AID, the DNA structural requirements of AID are important to consider next.

AID-type breaks at AID hotspot motifs

AID-type breaks occur at AID hotspot motifs, including WGCW, WRC, and CGC, specifically at cytosines of ssDNA (Pham et al. 2003; Yu et al. 2004). It is the most common type of breakage in the oncogenes, including BCL2, BCL1, E2A, MALT1, and CRLF2 in early B cells and BCL6 and MYC in mature B cells, involved in B cell translocations. The IGH locus is the common translocation partner of the B cell oncogenes.

In early B cell translocations, the RAG-type/AID-type breakage is commonly observed (Figure 2). The RAG motif found near the breakage sites at the IGH locus indicates that the breaks in IGH most likely arise from the aberrant V(D)J recombination events in these translocations. The breakpoints in B cell oncogenes in BCL2-IGH, BCL1-IGH, MALT1-IGH, and CRLF2-IGH translocations are in significant proximity to the CG motif and AID hotspot motifs (Table 2), suggesting the involvement of AID in the breakage phase. Again, this feature is not observed in T-ALL translocations because of the exclusive expression of AID in B cells but not T cells (Lieber et al. 2010; Lieber 2016).

Table 2.

Statistical analyses of sequence motifs in the proximity of fragile region breakpoints in patients.

DNA motif		CG			CGC			WRC			WGCW			CCCC
Translocation type	Fragile zone	Binomial	t-Test	U-test	Binomial	t-Test	U-test	Binomial	t-Test	U-test	Binomial	t-Test	U-test	Binomial	t-Test	U-test
E2A-PBX1	23 bp E2A fragile zone	8.7E-08	3.1E-06	8.3E-06	2.0E-12	8.6E-07	4.1E-06	4.2E-05	5.8E-04	1.1E-03	1.0E + 00	5.2E-05	1.0E + 00	1.0E + 00	3.7E-07	1.5E-09
E2A-HLF	23 bp E2A fragile zone	1.3E-03	1.7E-02	6.9E-03	6.9E-03	2.6E-02	1.5E-02	6.9E-03	3.3E-02	2.4E-02	1.0E + 00	1.8E-01	5.7E-01	1.0E + 00	5.6E-03	1.1E-03
BCL2-IGH	175 bp MBR	1.8E-96	1.5E-46	1.2E-42	4.1E-92	4.1E-52	1.4E-48	3.5E-10	1.2E-01	1.6E-01	1.4E-05	3.1E-04	1.1E-01	3.8E-09	1.9E-02	1.7E-01
	105 bp MCR	1.5E-08	1.1E-03	8.2E-05	1.0E + 00	2.1E-02	1.0E + 00	1.0E + 00	2.2E-01	1.2E-01	1.0E + 00	1.2E-01	9.3E-01	1.0E + 00	8.2E-03	1.0E + 00
	561 bp ICR	5.4E-18	2.7E-10	7.6E-13	2.1E-17	1.7E-09	4.8E-10	6.0E-06	6.8E-05	4.6E-06	5.3E-01	8.5E-05	1.1E-06	1.0E + 00	3.6E-03	1.0E + 00
BCL1-IGH	150 bp MTC	7.0E-09	1.1E-11	1.1E-12	1.5E-04	8.8E-09	7.0E-09	2.2E-02	1.2E-01	1.5E-01	1.0E + 00	1.1E-02	9.1E-01	9.6E-01	1.1E-05	6.0E-10
MALT1-IGH	86 bp fragile zone	3.3E-03	8.8E-03	6.2E-03	1.0E + 00	2.5E-01	5.7E-01	9.2E-01	4.7E-01	4.8E-01	5.6E-04	1.5E-03	1.6E-03	1.0E + 00	4.4E-01	3.6E-01
BCL6-IG	2156 bp break zone	3.5E-01	7.2E-02	5.6E-02	2.1E-01	1.5E-02	1.6E-02	2.2E-02	1.2E-03	1.2E-03	2.4E-02	2.9E-06	4.7E-07	2.8E-01	2.7E-01	3.2E-01
BCL6-non IG	2156 bp break zone	1.7E-03	8.3E-04	6.7E-04	4.9E-03	3.7E-04	4.6E-04	9.1E-04	2.1E-02	3.0E-02	2.6E-02	1.2E-02	2.3E-02	9.1E-01	4.7E-01	3.6E-01
MYC-IGH	4.1 kb break zone	6.9E-01	2.7E-01	7.7E-01	9.2E-01	1.1E-01	8.8E-01	2.9E-02	8.7E-02	1.0E-01	3.7E-03	2.6E-06	6.7E-06	6.5E-01	1.2E-02	9.9E-01

Our statistical analyses for all breakpoints in different B cell fragile zones in proximity to CG, AID hotspot motifs (CGC, WRC, WGCW), and CCCC motif (C-strings) are shown in the table (W = A or T; R = A or G). The statistical analyses were performed in the same way as described before using the Lieber Lab database (Tsai et al. 2008). The binomial statistic gives the probability that the E2A breakpoints occur precisely at the tested DNA motif by random chance. To test for proximity significance, rather than breakage precisely at a specific motif, the Student’s t-test and the Mann–Whitney U-test are used. DNA motifs to which the breakpoints show statistically significant proximity are highlighted in red for each fragile zone involved in different translocations. The breakpoints in BCL2 fragile zones, BCL1 MTC, E2A 23 bp fragile zone, and 86 bp MALT1 fragile zone are in statistically significant proximity to both CG and AID hotspot motifs. All events mentioned above except MALT1 breakpoints show significant proximity to C-strings. The BCL6 and MYC patient breakpoints are in statistically significant proximity to AID motifs.

Besides the common IGH translocation partner loci, other oncogenes can also participate in the early B cell translocations. E2A-HLF and E2A-PBX1 translocations are typical of the AID-type/uncertain-type category. The breakpoints of E2A gene in both translocations are in significant proximity to the CG motif and AID motifs (Table 2), suggesting the key role of AID in the initiation of the E2A breakage. In contrast to the DSBs at E2A, the breakpoints on PBX1 and HLF genes are randomly distributed within broad zones (230 kb in PBX1 intron 2 and 5.3 kb HLF intron 3) with no obvious clustering (Figure S7). Due to their largely random distribution within a given intron, the breakpoints of PBX1 and HLF are most likely random and due to the pathologic causes mentioned earlier. The growth advantage from the resulting fusion protein is likely the determining factor for which introns within PBX1 and HLF genes are involved in the translocations.

In mature B cells, AID-type/AID-type breakage accounts for most translocation events, including BCL6-IGH translocation and MYC-IGH translocation. The AID motif found around the breakpoints of IGH locus suggests an event due to CSR or SHM in the mature B cells. The significant proximity of BCL6 and MYC breakpoints to the CG motif and AID motifs also indicates that AID is involved in the breakage of those genes (Table 2). Given this, we next discuss recent data about the mechanisms underlying the breakage of oncogenes involved in B cell translocations.

Factors relevant to human early B cell translocations

Considering the time window and sequence signatures of the fragile regions, breakage events at BCL2, BCL1, E2A, MALT1, and CRLF2 are likely caused by similar mechanisms. First, the fragile regions of those genes are short and vary between 20 and 600 bp (Tsai et al. 2008; Greisman et al. 2012; Cui et al. 2013; Lieber 2016; Pannunzio and Lieber 2017) (Figure 3). Most of the breakpoints are within these focal regions, which are up to 400-fold more fragile compared with the adjacent and surrounding sequences (Tsai et al. 2008; Liu et al. 2021). For instance, among the 49 patients with E2A-PBX1 translocations, 36 of them have E2A breakage in the 23 bp fragile zone and 13 of them have breakpoints located outside of the 23 bp region but still within the 3.3 kb E2A intron 16. Based on the average patient number per base pair, the 23 bp E2A fragile zone is 426-fold more fragile than the other regions of the same intron. Second, breakpoints within these fragile zones are in proximity to the CG motif and AID hotspot motifs with high statistical significance (Table 2). Third, all the breakage events arise from the pro-/pre-B cell stage. The presence of N-nts in most of the junctional sequences is consistent with TdT activity, which is only expressed at pro-B/pre-B/pre-T cell stages (Supplementary Text). Fourth, the fragile zones tend to be in C-string-rich regions (Table 2; Figure S8). DNA with C-strings is in a B/A intermediate (non-B) DNA structure which undergoes more thermal fluctuation than DNA of random sequence and causes transcriptional pausing (see below) that would favor the action of AID (Dornberger et al. 1999; Trantírek et al. 2000; Tsai et al. 2009) (Figure 4). Fifth, no currently known common features, including histone modifications, transcriptional pausing, gene splicing, or replication origin proximity, are specific to all these fragile regions (Figure S7).

Figure 4.

Models for DNA breakage in early B cells. (A) Schematic models for the DNA breakage in early B cells. The fragile zones and their nearby regions usually contain a high density of C-strings that predispose the regions to a B/A intermediate (non-B) DNA structure. The DNA in the non-B structure tends to show increased thermal fluctuation that predisposes the fragile zones to AID activity without transcription. Frequent transcription through these fragile zones can increase their single-stranded character. RNA polymerase tends to move slowly and accumulates in regions adopting a non-B DNA structure, which further increases the duration of ssDNA. The cytosines in regions of ssDNA due to the transcription through these non-B regions are preferred substrates of AID deamination activity when they are in AID hotspot motifs. A long-lived DNA lesion can result if the cytosines are within a WRCG motif and in a methylated state. The persistent DNA lesions can be subject to nuclease attack inside the cells and then DSBs. (B) Steps leading to DNA breakage in fragile zones involved in early B cell translocations. The high density of C-strings around the fragile zones predisposes the fragile regions to a B/A intermediate DNA structure and then increased single-stranded character. Slowed RNA polymerase during transcription in regions adopting a non-B DNA structure and slippage between DNA direct repeats further increase their single-stranded character. The cytosines in regions of ssDNA are preferred sites of AID deamination activity when they are in AID hotspot motifs. As in panel A, a long-lived DNA lesion can result if the cytosines are within a WRCG motif in a methylated state, which is vulnerable to nuclease attack inside the cells.

AID-type breaks account for nearly half of breakpoints involved in translocations of all human hematopoietic malignancies (Tsai et al. 2008; Tsai, Lu, et al. 2010; Tsai, Yoda, et al. 2010; Greisman et al. 2012; Lu et al. 2013; Pannunzio and Lieber 2017; Liu et al. 2021) [see the red portion in Figure 5, and note the much smaller fraction that is RAG-type/RAG-type category (blue)]. We have reported the critical role of AID in the breakage phase of the fragile zones that are 20–600 bp in size for oncogenes of the early B cell stage (Tsai et al. 2008; Greisman et al. 2012; Cui et al. 2013; Pannunzio and Lieber 2017). The central question that has remained is what factors restrict all the DNA breakage to such small regions. As mentioned above, several features are commonly observed in the fragile zones of B cells, which may contribute to the fragility of those regions (Figure 4). First, the breakpoints of the oncogenes involved in B cell translocations are in significant proximity to C-strings (Table 2). A large percentage of them are predominantly located downstream of the fragile zones (Figure S8). For instance, the 23 bp E2A fragile zone is located at the upstream edge of a region containing the highest C-string density in the E2A intron 16 (Figure S11(A)). DNA with C-strings can adopt a non-B structure (Tsai et al. 2009), which predisposes the regions with increased thermal fluctuation which would increase the transient single-strandedness. Since AID can only deaminate cytosines in regions of ssDNA (Bransteitter et al. 2003; Pham et al. 2003), the increased single-stranded state of those oncogenes due to the high C-string density is one of the primary factors that limits the boundaries of the B cell fragile zones (Liu et al. 2021). Besides the E2A fragile zone, the single-stranded character of BCL1 MTC and BCL2 MBR is also reported in our previous studies (Raghavan et al. 2004; Raghavan et al. 2005; Tsai et al. 2009; Liu et al. 2021).

Figure 5.

Fraction of human hematopoietic malignancies explained by AID-type breaks. Each human hematopoietic malignancy is shown to reflect the fraction of all hematopoietic malignancies. The numbers of events reflect the incidence of all events per year in the United States, including all ages (adults and children) (Teras et al. 2016; Cancer Facts & Figures 2021). The translocations and their percentage in each malignancy are estimated from Swerdlow et al. (2008). The portion of translocations that involve at least one AID-type event, often at the oncogene, are shown within the red. Some are AID-type/AID-type events, such as translocations between BCL6 or MYC and the IGH switch regions. But most are AID-type/RAG-type events, such as BCL1, BCL2, MALT1, or CRLF2, to the IGH locus during failed D_H to J_H joining. The small fraction colored in blue is RAG-type/RAG-type events. Note that many other translocations are not included [for example, BCR–ABL1 translocations also occur in some B cell lineage acute lymphoblastic leukemia (ALL); and mixed-lineage leukemia (MLL)–AF9 occurs in some cases of acute myeloid leukemia (AML)] as this figure is not intended to be comprehensive; readers are referred to the current American Cancer Society website or the current version of the WHO text cited here (Swerdlow et al. 2008) for further details. CCND1: cyclin D1; CDKN1A: cyclin-dependent kinase inhibitor 1 A; ETO: eight twenty one protein; HOX11: homeobox; NPM1: nucleophosmin; PBX: pre B cell leukemia homeobox 1; PML: promyelocytic leukemia; RARA: retinoic acid receptor-α; SIL: SCL interrupting locus; TAF8: TATA-box binding protein associated factor 8; TCR: T cell receptor; WHSC1: Wolf–Hirschhorn syndrome candidate 1; WWOX: WW domain-containing oxidoreductase.

Second, most of the fragile zones are under active transcription in B cells, either to mRNA or lncRNA (Figure S9). The non-transcribed strand (NTS) within the transcription bubble is in a single-stranded state which is vulnerable to AID activity. When the fragile zones are transcribed (either to lncRNA for MALT1, or to both RNA and lncRNA for BCL2, E2A, MYC, and BCL6), the C-strings located in close proximity to the fragile zones can slow transcription transit time and therefore further increase access by AID. It has been shown that RNA polymerase II stalling sites are preferred by AID in class switch recombination (Pavri et al. 2010). The pausing of RNA polymerase during the elongation phase is sequence-dependent and occurs at GC-rich regions, with 65% of the pause sites at cytosines (Krohn and Wagner 1996; Watts et al. 2019). The pausing of RNA polymerase II was reported to correlate with G-strings on the NTS (Eddy et al. 2011). A C-string of 6 bp in size on the transcribed strand (TS) was found to be a strong transcriptional pause site (Pham et al. 2019). Pausing of RNA polymerase at the C-string-rich region may potentially increase the single-stranded duration of the NTS, which leads to increased AID activity. It has been reported that G-rich DNA regions on the NTS (C-rich on the TS) in proto-oncogenes are preferred sites of hypermutation and translocation in AID-positive B cell lymphomas (Duquette et al. 2007). Our study using wild-type E2A sequence and E2A sequence with disrupted C-strings indicates an increased AID targeting to the 23 bp E2A fragile zone due to C-strings located near and downstream of the fragile zone (Liu et al. 2021).

Third, DNA direct repeats are commonly found around the fragile zones (Figure S10). During transcription, the abundant DNA repeats within the fragile zones may lead to transient DNA slippage between, for example, the second repeat on the NTS and the first repeat on the TS. The exposed cytosines due to misalignment may also contribute to the frequent targets of AID activity in the fragile zones. A 6 bp DNA direct repeat flanks the 23 bp E2A fragile zone (Figure S11(A)). Disruption of this repeat is accompanied by significantly decreased AID activity within the fragile zone in a defined biochemical system (Liu et al. 2021).

Fourth, CpG sites within all B cell fragile zones have a detectable level of methylation (DNA feature panel in Figure S9). The methylated CG motif at AID hotspot sequences further narrows down the region of fragility. In regions with transient single-strandedness, all cytosines in AID hotspot motifs are targets of AID deamination. In comparison to the easily recognized and repaired U:G mismatch (Schmutte et al. 1995; Walsh and Xu 2006), the T:G mismatch generated by AID in presence of methylated cytosine at the CpG site is more persistent and has a higher chance to be converted to a DSB (Tsai et al. 2008; Cui et al. 2013). Studies on the other two human fragile regions (BCL1 MTC and BCL2 MBR) using a yeast model organism emphasize the importance of long-lived DNA lesions for DSBs (Pannunzio and Lieber 2017). The CpG sites within the fragile zones of B cells, where most of the patient breakpoints are clustered around, are frequently within WRCG motifs, suggesting the key role of DNA methylation in the breakage phase. Of note, the two CpG sites within the E2A fragile zone are both in the WRCG motif; moreover, they are the only two WRCG motifs that overlap with each other within the region of high C-string density (Figure S11(B)). Studies have shown that overlapping AID hotspot motifs are a critical factor in AID targeting (Han et al. 2011; Wei et al. 2015).

Mechanistic model for the clustered DNA breakage in human early B cell translocations

Based on recent mechanistic progress, we have the following model for why the translocations in the most common B cell malignancies occur in very restricted zones of only 20–600 bp (Figure 4). As mentioned above, C-strings within and near the fragile zones lead to increased thermal fluctuation of the fragile zones, which increases the accessibility of the fragile zones to AID. During transcription, the C-strings function as roadblocks to RNA polymerase, which contributes to the transcriptional pausing and accumulation of RNA polymerase within the fragile zones. The slowed transcription bubbles prolong the duration of the NTS in its single-stranded state, and the single-strandedness makes it vulnerable to AID activity. Transcription also increases the chance of DNA slippage between direct repeat sequences within and across the fragile zones, which form secondary structures that configure the fragile zones in a transient single-stranded state. Cytosines within the fragile zones in transient and intermittent single-stranded states are preferred sites of AID deamination, especially cytosines within AID hotspot motifs. In cases of methylated cytosines at CpG sites within AID hotspot motifs (i.e. the WRCG motif), AID can result in longer-lived DNA lesions, which can be converted to DSBs by common nucleases in all living cells (e.g. APE1) or B cell-specific nucleases (RAG or activated Artemis) (Ma et al. 2002; Tsai et al. 2008). Of note, the initial AID lesions generated within the fragile zone and its downstream region are likely to cause additional RNA polymerase pausing, thereby increasing ssDNA exposure. This would lead to an increase in AID deamination activity within this region (Gonzalez et al. 2020).

Our model (RAG or activated Artemis cutting at AID-deaminated methylated CpG sites in regions with increased single-stranded character due to non-B DNA structure or transcription pausing) explains the developmental stage, lymphoid lineage, and sequence specificity of the major human lymphoid translocation breakpoints.

Medical and clinical relevance of understanding the molecular mechanism of lymphoid translocations

Clinical relevance of the etiology

Understandably, many patients and their physicians may wonder whether they could have done something different to avoid getting a lymphoid malignancy. Understanding that AID action at CpG sites is a natural initial event for most patient lymphoid translocations can be reassuring that this is not necessarily due to a strong environmental or lifestyle factor. The risk of AID action in vertebrate B cells is the price paid for the benefit of AID contributing to the diversification of the immune system. We know that humans and mice lacking AID die of pneumonia at a very early age. Nevertheless, small environmental effects may exist. For example, bacterial and viral infections during the normal course of life appear to increase the likelihood of acute lymphoblastic leukemia (ALL) through a combination of effects that may prolong or increase the exposure of B cells to AID (Greaves 2018). However, such infections are intrinsic to nature.

Pathogenesis, prognosis, and diagnosis

By understanding the molecular mechanism of the translocations, physicians and their patients can understand several aspects relevant to the course of their disease. First, for mature B cell malignant clones (most of which bear translocations that arose when the B cells were at the pro-B/pre-B cell stage), the window during which AID and RAG enzymes were expressed is long past by the time that specific clone becomes clinically apparent. This means that they do not need to worry about ongoing additional cancers being related to the one they already have. Second, these are low-frequency events. There are millions of RAG and AID events per day, and the fraction that results in translocation is likely <1 per decade of life. Third, for immature B cell malignancies (e.g. ALL), there is usually ongoing RAG expression, which can result in additional genomic changes in the ALL, but these typically remain subclinical (Greaves 2018).

For diagnostic purposes, the molecular mechanism explains why screening for the breakpoint in the most common B cell malignancies can be targeted primarily, but not exclusively, to the observed fragile zones. If additional areas are of interest, then the molecular mechanisms described here will be helpful in deciding which areas are most likely to fit the key factors discussed for AID action in this review, specifically methylated CpG sites, flanked by C-strings and short direct repeats in the nearby sequence, with transcription providing the added probability of being a site of breakage.

Relevance of the mechanistic insights at lymphoid fragile zones to non-lymphoid malignancies

The joining phase of non-lymphoid translocations is the same as for lymphoid translocations. Most of the joining is done by NHEJ, with a smaller percentage of joining done by alternative end-joining (aEJ) (Ramsden and Nussenzweig 2021).

We discussed earlier that the breakage events in non-lymphoid cells can have pathologic causes (Figure 1(A)). The pathologic causes are much less constrained to small regions in the case of non-lymphoid malignancies. Because of this, the breakage zones for non-lymphoid cells are typically hundreds to thousands of times larger than the focal fragile zones that we have discussed here for lymphoid cells. Some fragile zones are called Fra sites (Glover et al. 2017). The latter can be millions of base pairs in length and are induced by aphidocholin in cells in culture. These are only rarely sites of spontaneous chromosomal breakage in vivo.

There is one particular non-lymphoid translocation that occurs in a very narrow fragile zone in patients who have previously been treated with the topoisomerase II inhibitor, mitoxantrone or epirubicin, to result in a t(15;17) translocation, which then results in a secondary malignancy, acute promyelocytic leukemia (Mays et al. 2010). The basis for this fragile zone is interesting but does not appear to be related to the factors that result in the 20–600 bp fragile zones discussed in this review.

Future questions

Influence of DNA repeats on DNA breakage

DNA repeats, including direct and inverted repeats, are commonly found in the fragile zones in B cell translocations. The strand slippage between direct repeats can contribute to the prolonged duration of single-stranded DNA which is vulnerable to AID activity. We do not know the exact effect of DNA repeats on the fragility of the translocation hotspots and how the length of the repeat, positions, sequences, and interspace regions can affect the AID activity within the fragile zones. Further studies are needed to experimentally examine these possibilities.

C-strings, DNA structure, and DNA breakage

We have found that C-strings are important for AID activity within the E2A fragile zones. The influence of the C-strings on the fragility of other fragile zones in BCL1, BCL2, MALT1, and others needs to be investigated. Furthermore, the impact of position, strand distribution, the density of the C-strings within and around the fragile zones must be addressed in the future.

Other factors affecting DNA breakage

Transcription, ncRNA, alternative splice sites, or other factors within non-coding regions that could be important for the clustered DNA breakage are possible directions for future studies of the targeting of AID to fragile zones.

Concluding comments

Most human B cell lymphomas have chromosomal translocations involving one break at an oncogene and a second DNA break at the immunoglobulin locus. In general, RAG-type/AID-type combinations are most frequent, such as at the BCL2, BCL1, CRLF2, and MALT1. The breakpoints of the oncogenes involved in early B cell translocations are highly focused to 20–600 bp regions, centering around the CG motif. We have reviewed the features at the DNA sequence and structural level around the fragile zones to understand the factors that may play key roles in the breakage phase.