Genetic interaction between the non‐homologous end‐joining factors during B and T lymphocyte development: In vivo mouse models

Non‐homologous end joining (NHEJ) is the main DNA repair mechanism for the repair of double‐strand breaks (DSBs) throughout the course of the cell cycle. DSBs are generated in developing B and T lymphocytes during V(D)J recombination to increase the repertoire of B and T cell receptors. DSBs are also generated during the class switch recombination (CSR) process in mature B lymphocytes, providing distinct effector functions of antibody heavy chain constant regions. Thus, NHEJ is important for both V(D)J recombination and CSR. NHEJ comprises core Ku70 and Ku80 subunits that form the Ku heterodimer, which binds DSBs and promotes the recruitment of accessory factors (e.g., DNA‐PKcs, Artemis, PAXX, MRI) and downstream core factors (XLF, Lig4 and XRCC4). In recent decades, new NHEJ proteins have been reported, increasing complexity of this molecular pathway. Numerous in vivo mouse models have been generated and characterized to identify the interplay of NHEJ factors and their role in development of adaptive immune system. This review summarizes the currently available mouse models lacking one or several NHEJ factors, with a particular focus on early B cell development. We also underline genetic interactions and redundancy in the NHEJ pathway in mice.

During the last years, several models have been proposed to explain how the two free DNA ends are brought back together through synapsis. In particular, Loparo's group 7 suggested a two-stage model of NHEJ synaptic complex assembly, where DNA ends are initially tied in a long-range complex, followed by transition into a short-range complex. In this model, Ku70, Ku80 and DNA-PKcs (DNA-PK) first participate in the formation of the initial long-range complex, where DNA ends are held sufficiently distant. Then, the short-range complex is formed by DNA-PK, XLF, Lig4 and XRCC4. PAXX and MRI have not been implicated in this model because their functions have not been identified. 7 Another model proposed by Lieber's group 24 suggests that there are two major structural complexes formed during the NHEJ synapsis. Ku70, Ku80, XRCC4 and Lig4 form the flexible synaptic (FS) complex, where XRCC4 and Lig4 bind to each DSB through interaction of Lig4 with Ku heterodimer. Subsequently, DSBs are brought together through interaction of XRCC4, giving rise to two Ku-XRCC4-Lig4-DNA complexes. XLF and PAXX both promote transition from FS to the second synaptic complex, called close synapsis (CS), although XLF stabilizes CS to a greater extent. XLF stimulates and impacts the general synapsis efficiency mediated by Ku-XRCC4-Lig4. It is suggested that DNA-PKcs is not required for the formation of either FS or CS. This latter model explains the evolutionarily central synaptic role of the core NHEJ factors, Ku70, Ku80, XRCC4, Lig4 and XLF. 24 For practical purposes, NHEJ can be divided into three major stages: DSB recognition, stabilization-processing and end ligation. 25 Initially, DSBs are recognized by the heterodimer Ku, which is formed by Ku70 and Ku80. Ku assists the recruitment of DNA-PKcs, [13][14][15][16] forming the DNA-PK holoenzyme. Subsequently, Artemis nuclease, 17 PAXX 18-21 and MRI 22,23 are recruited to the DSB sites. Finally, XLF, XRCC4 and Lig4 mediate the Ku-dependent DNA end ligation. 1 During the early stages of B and T cell development, NHEJ is required for the V(D)J recombination assembling immunoglobulin (Ig) and T cell receptor (TCR) genes using V, D and J gene segments. Both Ig and TCR provide antigen-binding specificity required for an efficient immune response. The proteins encoded by the recombination activating genes 1 and 2 (RAG1,2) form an endonuclease (RAG) that recognizes recombination signal sequences (RSSs) flanking the V, D and J gene segments. 26 Class switch recombination takes place in mature B cells, when constant regions of immunoglobulins switch from IgM to IgG, IgA or IgE. Immunoglobulins, or antibodies, play a crucial role in immune response through their effector functions. CSR is initiated by activation-induced cytidine deaminase (AID). In repetitive switch regions of Igh gene, AID deaminates deoxycytidine resulting in deoxyuracil (dC > dU). The dUs are excised by the uracil DNA N-glycosylase (UNG) enzyme, leaving an abasic (apyrimidinic/apurinic [AP]) site. 27 The AP sites are cut by AP endonuclease (APE)1 or APE2, producing DNA single-strand breaks (SSB). Two SSBs on the opposite DNA strands form DSBs, initiating the CSR. 27 The joining of DSBs during the CSR is performed both by the classical NHEJ (C-NHEJ) and alternative end joining (A-EJ). 28 Strikingly, the A-EJ can maintain up to 50% of CSR activity in the absence of core C-NHEJ factors, such as Ku70, XRCC4, Lig4 and XLF. 1 Other proteins, such as members of the DNA polymerase X family and terminal deoxynucleotidyl transferase enzyme (TdT), can also be involved in C-NHEJ within B cells. 29,30 For example, DNA polymerase proteins Pol λ and Pol μ promote DNA end joining through the processing of DSBs, while TdT increases the antibody and TCR repertoire by adding non-template nucleotides prior to ligation of DNA ends during V(D)J recombination. 29,30 Proteins such as nipped-Blike protein (NIPBL) and breast cancer 1 (BRCA1) have also been shown to play a role in NHEJ. 31,32 | 3 of 9 CASTAÑEDA-ZEGARRA ET Al. homozygous mice. Later, transgenic mouse models deficient for Dna-pkcs gene were generated by several groups. [13][14][15][16] DNA-PKcs-deficient mice (Dna-pkcs −/− ) are live-born and possess a SCID phenotype due to inefficient coding-end (CE) joining during the V(D)J recombination 13 (Figure 1). Artemis −/− mice have also been observed to exhibit a SCID phenotype due to lack of CE joining. 17 Thus, both DNA-PKcs and Artemis are required for processing of RAG-mediated hairpin-sealed DNA ends (CEs) during V(D)J recombination, although repair of blunt signal ends (SEs) remains efficient in mice lacking Artemis 17 or DNA-PKcs. [13][14][15] Inactivation of Ku70 11 or Ku80 10 in mice results in reduced body weight and a SCID phenotype. Lack of B and T lymphocytes in Ku70 −/− and Ku80 −/− mice is explained by inefficient joining of both RAG-induced blunt SEs and hairpin-sealed CEs. 10,11 In contrast, inactivation of Lig4 8 or Xrcc4 9 results in embryonic lethality in mice, presenting challenges for in vivo studies. However, cell studies show that inactivation of Lig4 or Xrcc4 led to inefficient joining of both SEs and CEs, resembling Ku-deficient phenotypes in mice. This suggests that such in vivo models would yield an immunodeficient animal due to ablated V(D)J recombination and lack of B and T cells, if one could be generated. 8,9 Several single-deficient mouse models initially suggested that XLF, PAXX and MRI are dispensable for the V(D)J recombination. Particularly, mice lacking XLF/Cernunnos 34,35 possess both mature B and T cells, despite being characterized by modest lymphocytopenia, and reduced repertoires of B cell receptors and TCRs. 34,35 Xlf −/− lymphocytes support efficient V(D)J recombination in vitro, including both SE and CE repair. 34,35 Mice lacking either PAXX 18,19,21,36 or MRI 22,23 possess normal counts of mature B and T cells, efficiently supporting both SE and CE repair during the V(D)J recombination. However, more complex mouse models have revealed that XLF, PAXX and MRI are required for V(D)J recombination, although their functions are compensated by each other and additional proteins due to the extensive genetic interaction inside the NHEJ pathway, as well as interaction between the NHEJ and DDR pathways. 1,37 Xlf −/−34, 35 and Dna-pkcs −/−13 mice show notable radiosensitivity, but result in viable mice. However, double-deficient Xlf −/− Dna-pkcs −/− mice are characterized by perinatal lethality and increased genomic instability, due to nearly no NHEJ. 38 Genetic interaction No genetic interaction Single deficiency NHEJ factors in vivo (Figure 1). Mice lacking both XLF and Artemis are live-born, are fertile and have a SCID phenotype resembling single-deficient Artemis −/− mice, 38 suggesting no genetic interaction between Xlf and Artemis in vivo. Later, genetic interaction between Xlf and Paxx was characterized in vivo independently by four research groups. 18,19,36,39 Mice lacking both XLF and PAXX possess late embryonic lethality, increased genomic instability and immunodeficiency; and Xlf −/− Paxx −/− lymphocytes are unable to sustain repair of both SEs and CEs generated during V(D)J recombination. 18,19,36,41 The embryonic lethality of Xlf −/− Paxx −/− mice is p53-dependent, and both Xlf −/− Paxx −/− Trp53 +/− and Xlf −/− Paxx −/− Trp53 −/− mice are live-born, but possess significant weight reduction and a leaky SCID phenotype with nearly no mature B and T lymphocytes. 39,42 Furthermore, Xlf −/− Mri −/− double knockout mice are embryonic lethal, 23 but can be rescued by inactivation of one or two alleles of Trp53. 42 Mice lacking both XLF and MRI are characterized by leaky SCID with nearly no mature B and T lymphocytes due to the V(D)J recombination defect. 23 In particular, the lymphocytes lacking both XLF and MRI are unable to efficiently ligate both RAG-mediated SEs and CEs in vitro. 23 Both the double-deficient mouse model lacking XLF and PAXX and the model lacking XLF and MRI are characterized by leaky SCID with low but detectable levels of mature B and T lymphocytes. 42 This phenotype is likely possible due to residual NHEJ activity in developing Xlf −/− Paxx −/− and Xlf −/− Mri −/− lymphocytes in vivo. 39,42 Thus, there is evidence that Xlf genetically interacts with Dna-pkcs, [38][39][40]19,36,39,42 and Mri. 23 45 Moreover, Lig4 −/− Trp53 +/− mice present a more severe phenotype than Lig4 −/− Trp53 −/− littermates, likely due to an incomplete block of DNA damage-induced apoptosis in the presence of one Trp53 allele. 45 In addition, inactivation of one or two alleles of Atm rescued lethality in Lig4-deficient mice. 46 Both Lig4 −/− Atm +/− and Lig4 −/− Atm −/− mice displayed impairment in lymphocytic development, growth retardation and short lifespan (up to 2 days postnatally). 46 Further, inactivation of one or two alleles of Trp53 rescues embryonic lethality of Xrcc4 −/− mice. 47 The Xrcc4 −/− Trp53 +/− mice possess a more severe phenotype when compared to the Xrcc4 −/− Trp53 −/− littermates, although both models are characterized by reduced body weight, increased genomic instability and SCID due to inability to repair RAG-induced DSBs in developing B and T cells. 47 Several complex mouse models have also revealed a lack of genetic interaction between different pairs of NHEJ genes (Figure 1). In particular, mice lacking

| Class switch recombination
Class switch recombination occurs in mature B cells following the efficient V(D)J recombination in vivo. However, several experimental models allow for the bypass of V(D) J recombination to determine the impact of specific factors on CSR, even though these factors are required for earlier stages of B cell development (e.g., Artemis, DNA-PKcs). Knocking-in pre-assembled heavy and light chains of the immunoglobulin gene ('HL') allows for the development of mature B cells in mice that otherwise lack the capacity for V(D)J recombination. 49 For example, lack of Artemis or DNA-PKcs 49 moderately affects the CSR in mature B cells. Additionally, it was found that CSR levels were reduced two-to threefold in cells lacking XLF, 34,35 Ku70, 44 Ku80, 44 Lig4 28 or XRCC4. 28 PAXX seems to be dispensable for CSR in wild-type cells, 18,19,21,36,50 while inactivation of Mri results in modest CSR defects. 22,23,42 DNA-PKcs and XLF are functionally redundant in CSR. 38 46 (Figure 2). Combined inactivation of Atm and Dna-pkcs in cells results in more severe CSR defects than in single-deficient controls. 52 Moreover, Atm −/− Dnapkcs −/− pro-B cells lack repair of both SEs and CEs during the attempted V(D)J recombination. 53 Furthermore, Xlf genetically interacts with Atm, 53bp1, H2ax and Mdc1. 37,[54][55][56] (Figure 2). Taken together, evidence shows that Atm genetically interacts with Ku70, Ku80, Dna-pkcs and Xlf, but not with Paxx or Mri. Correspondingly, Xlf genetically interacts with Atm, H2ax, Mdc1 and 53bp1.

Xlf and Rag
Xlf has also been shown to genetically interact with Rag2. 58 Mutation in the Rag2 gene results in the truncated protein 'core Rag2', which continues to support DSB formation and DNA repair in developing B and T lymphocytes. However, in XLF-deficient cells, this 'core Rag2' activity is lost, and V(D)J recombination does not proceed. This finding suggests a potential role for RAG in both the induction of DSBs and DNA repair, as the RAG complex supports tethering of DNA ends before ligation 58 (Figure 2).
Hence, deficiencies in NHEJ often result in neuronal apoptosis, likely due to accumulation of DSBs in post-mitotic neurons.

Genetic interaction No genetic interaction
Inactivation of p53 prevents neuronal apoptosis, for example, by allowing A-EJ to repair DSBs in NHEJ-deficient cells.

| NHEJ in mouse and human
Mutations in several NHEJ genes have been identified in humans. 61,62 For instance, patients with mutations in XLF, DNA-PKCS/PRKDC and LIG4 genes display severe clinical features, characterized mainly by SCID, delayed growth and neurological abnormalities. 61,62 In mice, XLF deficiencies lead to a modest lymphocytopenia and defect in CSR 34,35 ; DNA-PKcs deficiencies lead to a SCID phenotype but no neural complications 59 ; and Lig4-deficiencies are embryonic lethal. 8 In the same manner, ARTEMIS/DCLRE1C-deficient patients are characterized by SCID but not neurological defects 62 ; similar to Artemis-deficient mice. 17 On the other hand, Xrcc4 −/− mice are embryonic lethal, 9 unlike XRCC4deficient patients who only display neurological problems. 63 Mutations in several NHEJ genes have not yet been found in human immunodeficient patients up to the present. These genes include Ku70, Ku80, PAXX and MRI. Ku70 and Ku80 might be essential in human cells, and therefore, mutations in KU70 and KU80 genes might be identified only by analysing embryonic samples. However, mutations in accessory factor genes PAXX and MRI might present without clinical features, based on the knowledge we have obtained from mouse models. 18,19,[21][22][23]36 In the latter case, XLF might compensate for deficiencies in PAXX and MRI in human cells. Sometimes, dramatic differences in phenotypic presentation between mice and humans lacking the same NHEJ factor can be explained, for example, by minor sequence changes between species, resulting in significant changes in proteinprotein and protein-DNA interactions, and inability for other factors to compensate the protein loss.

| Potential reasons for genetic interactions between the NHEJ factors
There are several types of genetic interaction between the DNA repair factors and several potential explanations for them, although detailed mechanisms have not been elucidated yet. Why does inactivation of Ku70 11 or Ku80 10 result in viable mice, while inactivation of Lig4 8 or Xrcc4 9 results in embryonic lethality? This cannot simply be due to lack of NHEJ activity, because cells lacking any of these factors are characterized by similar genomic instability. 38,40,57 The Ku70/ Ku80 complex seems to be toxic for the cells when the NHEJ pathway is blocked due to defects in downstream factors. It is possible that Ku may block access to DSB sites from other DNA repair pathway proteins, preventing DNA ligation and eventually resulting in the accumulation of DSBs, activation of p53 and apoptosis. Interestingly, Ku-deficient cells rely on other DNA repair pathways, such as homologous recombination and alternative end joining. This could also explain why inactivation of Ku70 or Ku80 rescues embryonic lethality of Lig4 −/− mice. 43,46 Similarly, lack of Ku rescues embryonic lethality in mice lacking XLF/DNA-PKcs, 38,40 as well as in mice with an inactivating DNA-PKcs point mutation. 60 Following the same logic, one can predict that inactivation of Ku70 or Ku80 would also have the ability to rescue synthetic lethality between Xlf and Paxx, and between Xlf and Mri. We can also predict that inactivation of all NHEJ genes in a mouse would result in a phenotype similar to those of Ku70 −/− or Ku80 −/− mice, 22 and suggest that all NHEJ genes function in a purely Ku-dependent manner.
Synthetic lethality between Xlf and Dna-pkcs, [38][39][40] Xlf and Paxx 18,19,36,39,42 and Xlf and Mri 23,42 results in phenotypes similar to Lig4 −/− and XRCC4 −/− . In all these cases, it is likely that the DNA ligation step is impaired, while Ku70/Ku80 remains functional. There are several potential explanations for the functional redundancy observed between XLF and other factors in NHEJ and DDR. 1,38,54,56 First, XLF and the second factor could have identical functions, such as having a role in stabilizing the DNA repair complex. A second explanation could be that XLF and the second factor could have purely complementary functions; for example, one protein stimulates DNA ligation while another one is required for DNA end tethering. However, the question of XLF's functionally redundancy with so many other factors 1,18,19,23,36,[38][39][40]42,54 is still an enigma in the field of DNA repair.
ATM and DNA-PKcs are both protein kinases. Synthetic lethality between Atm and Dna-pkcs 46 is reasonable to predict because these two proteins can partially compensate for each other's activity when one is inhibited, but no other protein can compensate combined ATM/DNA-PKcs deficiency. 52,53,64 DNA-PKcs is part of the DNA-PK holoenzyme, which includes Ku70 and Ku80. Both Ku70 and Ku80 are synthetic lethal with Atm, 46 meaning that ATM is functionally redundant with the Ku70/Ku80/DNA-PKcs complex, and DNA-PKcs will likely be inactive in cells lacking Ku70 or Ku80.