An XRCC4 mutant mouse, a model for human X4 syndrome, reveals interplays with Xlf, PAXX, and ATM in lymphoid development.

We developed an Xrcc4M61R separation of function mouse line to overcome the embryonic lethality of Xrcc4-deficient mice. XRCC4M61R protein does not interact with Xlf, thus obliterating XRCC4-Xlf filament formation while preserving the ability to stabilize DNA ligase IV. X4M61R mice, which are DNA repair deficient, phenocopy the Nhej1-/- (known as Xlf -/-) setting with a minor impact on the development of the adaptive immune system. The core non-homologous end-joining (NHEJ) DNA repair factor XRCC4 is therefore not mandatory for V(D)J recombination aside from its role in stabilizing DNA ligase IV. In contrast, Xrcc4M61R mice crossed on Paxx-/-, Nhej1-/-, or Atm-/- backgrounds are severely immunocompromised, owing to aborted V(D)J recombination as in Xlf-Paxx and Xlf-Atm double Knock Out (DKO) settings. Furthermore, massive apoptosis of post-mitotic neurons causes embryonic lethality of Xrcc4M61R -Nhej1-/- double mutants. These in vivo results reveal new functional interplays between XRCC4 and PAXX, ATM and Xlf in mouse development and provide new insights into the understanding of the clinical manifestations of human XRCC4-deficient condition, in particular its absence of immune deficiency.

Introduction

Living organisms face DNA double-strand breaks (DSBs), the most toxic DNA lesions, from random or programmed (prDSBs) origins (Betermier et al., 2020), such as during the development of the adaptive immune system through V(D)J recombination. V(D)J recombination results in the somatic rearrangement of variable (V), diversity (D), and joining (J) elements of antigen receptor loci in T- and B-cell precursors (Jung et al., 2006). It is initiated by the domesticated transposase Recombination-Activating Genes 1 and 2 factors (RAG1/2), which introduce prDSBs at the border of V, D, and J elements within recombination signal sequences (RSS). The non-homologous end-joining (NHEJ) machinery is the sole DNA repair pathway to cope with these lymphoid-specific prDSBs. Briefly, the NHEJ pathway is signaled by the Ataxia-Telangiectasia mutated (ATM) protein (Arya and Bassing, 2017; Bassing et al., 2002) and prompted by the Ku70-Ku80 complex that recognizes broken DNA ends and recruits the DNA-dependent protein kinase-catalytic subunit (DNA-PKcs), which further activates the pathway (Sibanda et al., 2017). Broken DNA ends are processed by Artemis endo/exonuclease, which specifically opens the RAG1/2-generated hairpins at DNA ends during V(D)J recombination (Chang et al., 2017). DNA ligase IV (Lig4), X-ray repair cross-complementation 4 (XRCC4, or X4), XRCC4-like factor (Xlf, also known as Cernunnos), and Paralog of XRCC4 and Xlf (PAXX) (Craxton et al., 2015; Ochi et al., 2015; Xing et al., 2015) ensure ligation of broken ends (Lieber, 2010). Lig4 and XRCC4 are critical factors during development apart from their essential function during V(D)J recombination, and both Lig4-/- and Xrcc4-/- mice suffer late embryonic lethality caused by apoptosis of post-mitotic neurons (Barnes et al., 1998; Frank et al., 1998; Shull et al., 2009). Likewise, human patients with LIG4 syndrome present with dwarfism, microcephaly, and various degrees of immunodeficiency caused by hypomorphic mutations in the Lig4 gene (O’Driscoll et al., 2001; Staines Boone et al., 2018).

In most of the cases, defects in core NHEJ factors result in severe combined immunodeficiency (SCID), owing to aborted V(D)J recombination both in humans and animal models (de Villartay et al., 2003). NHEJ deficiency also results in genetic instability with the development of Pro-B cell lymphomas harboring chromosomal translocations when associated with Trp53-/- (Alt et al., 2013). Xlf/Cernunnos-deficient patients also present with developmental features including dwarfism, microcephaly, and combined immunodeficiency (CID) (Buck et al., 2006). However, V(D)J recombination is not severely affected nor in human (Buck et al., 2006; Recio et al., 2018; van der Burg and Gennery, 2011) or murine (Li et al., 2008; Roch et al., 2019; Vera et al., 2013) settings of Xlf deficiency. This paradoxical situation is a consequence of the functional redundancy between Xlf and several DNA repair factors, including PAXX, ATM, H2A.X, MDC1, MRI, and 53BP1 as revealed by the complete V(D)J recombination defect in combined deficient settings (Abramowski et al., 2018; Beck et al., 2019; Hung et al., 2018; Musilli et al., 2020; Oksenych et al., 2012; Zha et al., 2011) as well as RAG2 itself (Lescale et al., 2016a). Consistent with their overall efficient V(D)J recombination, Nhej1-Trp53 DKO mice indeed do not develop Pro-B cell lymphomas (Vera et al., 2013).

To account for this functional redundancy, we proposed a model in which prDSBs may benefit from evolutionary conserved DNA repair mechanisms as to avoid their intrinsic oncogenic potential (Betermier et al., 2020). In the particular context of V(D)J recombination, this mechanism would operate through a redundant ‘double DNA repair synapse,’ which strictly operates during V(D)J recombination to prevent genomic instability, but not in case of genotoxic-induced DNA damage (Abramowski et al., 2018; Lescale et al., 2016a). The ‘two-synapses’ model accounts for the absence of V(D)J recombination defect in the absence of Xlf. One essential actor of the ‘two-synapses’ model is the RAG2 factor itself, which together with RAG1 is known to remain on DNA broken ends during V(D)J recombination, forming the so-called post-cleavage complex (PCC) (Schatz and Swanson, 2011). We previously established that the C-terminus of RAG2 is determinant in complementing the lack of Xlf (Lescale et al., 2016a). Indeed, the combined absence of the C-terminus of RAG2 and Xlf results in SCID mice owing to a complete block of V(D)J recombination. In the ‘two-synapses’ model, the first synapse would be mediated by RAG2, PAXX, and ATM signaling as suggested by efficient V(D)J recombination in multiply deficient v-Abl transformed Pro-B cells (Lescale et al., 2016a; Lescale et al., 2016b);. The second synapse is constituted by the Xlf-X4 filament or the bona fide NHEJ core complexes, the structure of which was recently resolved through cryoelectron microscopy (cryo-EM) (Chaplin et al., 2021a; Chaplin et al., 2021b; Zhao et al., 2020). The V(D)J recombination-specific ‘two-synapses’ apparatus appears as a double-edged sword backup system to avoid genomic instability. Indeed, RAG2/Xlf double mutant mice develop typical NHEJ-deficient pro-B cell lymphomas when crossed onto a Trp53-/- background (Lescale et al., 2016a).

One question remains as to the contribution of X4 in this model given that classical KO strategies do not allow to directly assess the function of X4 since its absence results de facto in a complete Lig4 deficiency. Reminiscent to the Xlf-deficient condition, human patients identified with X4 mutations present DNA repair defect hallmarks, including dwarfism, microcephaly, increased cellular sensitivity to radiomimetic agents, but, strikingly, no immunodeficiency (de Villartay, 2015; Saito et al., 2016). The absence of noticeable immunophenotype argues for a dispensable role of X4 during V(D)J recombination, outside its general Lig4 stabilization function. It also interrogates the positioning of X4 within the ‘two-synapses’ model knowing its structural and functional relationship with Xlf (Callebaut et al., 2006).

XRCC4 is a keystone factor, playing two independent roles. First, X4 stabilizes Lig4 through its C-terminal coiled-coil domain (Grawunder et al., 1997). Indeed, Lig4 expression is abrogated in Xrcc4 Knock Out (KO) models both in vivo and in vitro (Gao et al., 1998). In addition, X4 and Xlf homodimers interact through their N-terminal globular head to form long polymeric ‘filaments’ (Reid et al., 2015; Ropars et al., 2011). X4-Xlf filaments generate a ‘DNA repair synapse’ to tether broken DNA ends (Brouwer et al., 2016; Reid et al., 2015). DNA-end synapsis is a central issue during NHEJ, and the recent development of in vitro single-molecule technologies has highlighted the dynamic formation of DNA end-to-end synapses (flexible/long range for DNA end tethering and close/short range for DNA ligation) in addition to the Xlf-X4 filament, in which the various core NHEJ DNA repair factors (Ku70/80, DNA-PKcs, Xlf, X4/L4, and PAXX) participate to various degrees, in particular the association of Xlf with both X4 and Ku (for a recent review, see Zhao et al., 2020). Several studies recently reported on the details of the structural assembly of these complexes using cryo-EM, thus improving our understanding on the composition of these complexes as well as the dynamics of the transition between various states during NHEJ-mediated DNA repair (Chaplin et al., 2021a; Chaplin et al., 2021b; Zhao et al., 2020). These studies support in particular the interaction of the L4X4 complex with that of Ku70/80 previously proposed by Costantini et al., 2007. The V(D)J recombination phenotype of Nhej1 KO mice would argue that the RAG2-mediated DNA tethering is also redundant with these DNA end-to-end synapses. Nevertheless, the intimate nature of DNA end joining during V(D)J recombination may not always strictly coincide with what we know for the repair of genotoxic DNA breaks, precisely because of the existence of the ‘two-synapses’ mechanism.

Since X4 is compulsory for Lig4 stabilization, Xrcc4-/- mice phenocopy Lig4-/- condition, with an embryonic lethality and SCID phenotype (Gao et al., 1998). The embryonic lethality is rescued on Trp53-/- background, but the SCID resulting from a complete block of lymphocyte development remains (Gao et al., 2000). Furthermore, Xrcc4-/-Trp53-/- DKO mice develop Pro-B cell lymphomas (Chen et al., 2016; Gao et al., 2000). To avoid disrupting the critical Lig4 stabilization function of X4, we engineered a Xrcc4 knock-in (KI) mouse model harboring the M61R missense mutation that abrogates X4-Xlf interaction (Ropars et al., 2011), while keeping the Lig4 interaction domain unperturbed. Xrcc4 mice are viable attesting for the stabilization of functional Lig4, thus allowing the study of lymphocyte development. Introduced on Atm-/-, Paxx-/-, and Nhej1-/- backgrounds, the Xrcc4 mutation allows to address the ‘two-synapses’ DNA repair model in V(D)J recombination, and to expand the picture of NHEJ apparatus in brain and lymphocyte development.

Results

Generation of X mice

X4 and Xlf interact through a hydrophobic interface, which is disrupted by the X4M61R substitution, thus resulting in the loss of X4-Xlf filament (Ropars et al., 2011). We developed an Xrcc4 KI mouse model through CRISPR/Cas9 (Figure 1A). Homozygous Xrcc4 mice (Xrcc4 mice) were viable arguing against any harmful impact of the M61R substitution during embryonic development, as opposed to the embryonic lethality of Xrcc4 KOs.

Figure 1.

X4 and Lig4 expression and impaired non-homologous end-joining (NHEJ) in Xrcc4 mice.

(A) Schematic representation of CRISPR/Cas9-driven homologous recombination strategy to generated Xrcc4 mouse model. (B) 5′UTR to 3′UTR Xrcc4 RT-PCR in mouse thymocyte extracts from four littermates. The lower transcript represents the splicing out of exon 3 as described (Gao et al., 1998). (C) WB analysis from mouse whole tissue extracts. (D) Western blot (WB) analysis from mouse embryonic fibroblast (MEF) extracts. (E) WB analysis from mouse whole tissue extracts. WB were performed at least three times with two animals per genotypes. (F) MEFs’ sensitivity to DNA double-strand break (DSB)-inducing agent phleomycin. Statistical tests for Xrcc4 vs. WT (gray *) and Xrcc4 vs. Xrcc4-/- (black *). (G) Relative survival of CD3/CD28-activated mature T cells following irradiation and 4 hr recovery. Statistical analysis for Xrcc4 vs. WT (gray *) and Xrcc4 vs. Nhej1-/- (green *).

(A) Sequence alignment of WT and upper/lower Xrcc4 transcripts (Figure 1B) showing M61R and silent mutations on the full-length (upper) form and the splicing out of exon 3 on the lower form. (B) Nucleotide sequence traces of upper and lower Xrcc4 transcripts.

X4M61R stabilizes Lig4

RT-PCR analysis of X4 mRNA expression unveiled two isoforms in M61R mutation-bearing cells (Figure 1B). The expression of the full-length transcript carrying the five mutations was reduced (Figure 1B, upper band) and was accompanied by a shorter isoform (Figure 1B, lower band). Sequencing this alternative transcript revealed the out-of-frame splicing-out of exon 3 (Figure 1—figure supplement 1). The same X4 transcript missing exon 3 was described in a previous Xrcc4 KO mouse model (Gao et al., 2000; Gao et al., 1998). A subsequent attempt to generate Xrcc4 KI mouse model using an ssODN template lacking the four silent mutations resulted in the same pattern of expression of the M61R allele (data not shown). At the protein level, X4 expression was barely detectable in spleen, brain, thymus, and mouse embryonic fibroblasts (MEFs) from Xrcc4 mice as opposed to their homozygous WT counterparts (Figure 1C and D). In summary, Xrcc4 mice express two X4 transcripts, one lacking exon 3 and resulting in a complete loss of function, and a second, full length, harboring the M61R missense variant encoding a weakly expressed X4 protein lacking its ability to interact with Xlf. Of note, Lig4 protein was readily detectable by WB in protein lysates from Xrcc4 mice spleen and thymus, attesting for its stabilization by the X4M61R protein (Figure 1E).

Figure 1—figure supplement 1.

Nucleotide sequence of two Xrcc4 transcripts in Xrcc4 mice.

(A) Sequence alignment of WT and upper/lower Xrcc4 transcripts (Figure 1B) showing M61R and silent mutations on the full-length (upper) form and the splicing out of exon 3 on the lower form. (B) Nucleotide sequence traces of upper and lower Xrcc4 transcripts.

We conclude that the Xrcc4 mouse model represents a hypomorphic condition in which the X4M61R protein, although weakly expressed, retains the capacity to stabilize Lig4 and thus ensures proper embryonic development with viable animals at birth.

DSB repair defect in X4M61R cells

MEFs from Xrcc4 mice presented a statistically significant decrease in cell viability compared to WT cells upon phleomycin, a DSB inducer, treatment, attesting for a profound DNA repair defect (Figure 1F). Likewise, splenic T cells from Xrcc4 mice showed an increased IR sensitivity with an almost complete loss of viability at 2 Gy (Figure 1G). These experiments demonstrate that the X4M61R mutant protein, which preserves the stabilization of Lig4, is nevertheless impaired in its capacity to achieve full repair of DSB introduced by random genotoxic agents.

Lymphocyte development in Xrcc4 mice phenocopies that of Nhej1-/- mice

We analyzed the impact of X4M61R substitution on T cell development in the thymus. Thymocyte counts were significantly reduced in Xrcc4 thymus (Figure 2A), but remained in the range of Nhej1-/- thymus cellularity (respectively [*106] 26.8 ± 2.6 SEM vs. 115 ± 15 SEM in WT, p<0.0001, vs. 12.5 ± 2.6 SEM in Nhej1-/-, ns). The thymocyte maturation program, identified through the various CD4/CD8 stages, was similar in Xrcc4 and WT littermates (Figure 2B), yet with a mild accumulation of CD44-CD25+ CD28- DN3A Xrcc4 thymocytes (respectively 81.8% ± 0.9 SEM vs. 76.1% ± 0.7 SEM, p<0.0001) (Figure 2C), a trait shared by thymocytes from Nhej1-/- mice (81.1% ± 1.4 SEM) as described (Roch et al., 2019). Purified Xrcc4 thymocytes presented an increased apoptosis (7AAD+/AnnV+), similar to that of Nhej1-/- thymocytes (respectively 47.4% ± 1.7 SEM vs. 33.2% ± 1.3 SEM for WT, p=0.0003, vs. 59.5% ± 1.6 SEM, for Nhej1-/-, ns) (Figure 2D and E). This was accompanied by the expression of TP53 target genes Cdkn1a, Bbc3, and Bax, in both Xrcc4 and Nhej1-/- thymocytes (Figure 2F). Lastly, TCRα repertoire was skewed in Xrcc4 thymocytes (Figure 2G), with the decreased usage of distal VαJα rearrangements, a hallmark of reduced thymocyte viability or suboptimal V(D)J recombination.

Figure 2.

Xrcc4 mice exhibit modest T cell development defect in the thymus phenocopying Nhej1-/- mice.

(A) Thymus cellularity of WT, Xrcc4, and Nhej1-/- (Xlf) 6–9-week-old mice. (B) Immunostaining of thymus cellular populations. Various relevant populations are highlighted in black gates. (C) Quantification of DN3A thymocyte subpopulation (DN3 CD28-CD25+) gated from total DN3 thymocytes (CD4-CD8-CD44-CD25+). (D) Thymocyte apoptosis analysis after 20 hr of culture. AnnexinV+/7AAD+ apoptotic cells are highlighted in black gates. (E) Quantification of thymocyte apoptosis (AnnexinV+/7AAD+ cells) after 20 hr of culture. (F) Quantitative RT-PCR analysis of TP53 target genes Cdkn1a (encoding P21), Bbc3 (encoding PUMA), and Bax in thymocyte extracts. There is no statistical difference between Xrcc4 and Nhej1-/- in the expression of these TP53 target genes. (G) Illustrative representation of principal component analysis/unsupervised hierarchical clustering (PCA/HC) analysis of mTRAJ-mTRAV combinations in thymus according to PROMIDISα. (H) Immunostaining of bone marrow cellular populations. V(D)J recombination-dependent developmental stages are highlighted within red rectangles. Various relevant populations are highlighted in black gates. Quantification of B-cell population (B220+) (I) and immature B-cell population (CD19+ B220+ IgM+) (J) from total bone marrow.

The proportion of mature B lymphocytes in the bone marrow (BM) was not significantly altered in Xrcc4 mice (Figure 2H, I), while significantly decreased in Nhej1-/- mice (respectively 21.7% ± 1.8 SEM vs. 26.0% ± 1.1 SEM, ns, vs. 13.3% ± 0.6 SEM, p=.008). However, the proportions of immature B cells were decreased in Xrcc4 mice (Figure 2J), close to the already described decline of Nhej1-/- immature B cells (respectively 11.3% ± 3.3 SEM vs. 19.3% ± 2.6 SEM, p=0.0002, vs. 8.7% ± 2.5 SEM, ns).

These results indicate that B- and T-lymphocyte development is not severely affected in Xrcc4 mice, despite the reduced expression of X4M61R. The minor accumulation of DN3A thymocytes and immature B cells suggests a modest decreased V(D)J recombination efficiency, supported by the bias in TCRα usage. This is in sharp contrast with Xrcc4-/- and Lig4-/- mice, which present a complete arrest of lymphocyte maturation and a SCID phenotype. We conclude that when Lig4 is stabilized, the contribution of X4 through Xlf/X4 interaction, disrupted in the context of X4M61R, is not critical for V(D)J recombination in vivo.

PAXX and ATM are compensatory factors for X4 in lymphoid development

We previously proposed a model of ‘double DNA repair synapse’ during V(D)J recombination according to which the X4-Xlf ‘filament’ and the RAG1/2 PCC, together with ATM and PAXX, provide two complementary means of DNA end synapsis (Abramowski et al., 2018; Betermier et al., 2020; Lescale et al., 2016a). To address the role of X4 in this model, we crossed Xrcc4 onto Paxx-/- and Atm-/- mice (hereafter denoted as Xrcc4 and Xrcc4). Both double mutant mice were viable but exhibited severe growth retardation and facial dysmorphia as a consequence of either the X4M61R mutation and/or the overall reduction in X4/L4 expression (Figure 3—figure supplement 1).

Figure 3—figure supplement 1.

Growth retardations of Xrcc4 and Xrcc4 mice.

(A–C) Photographs of 7-week-old littermates.

The thymus and spleen cellularity was severely reduced in both X4 and Xrcc4 mice (Figure 3A and B) as compared to their Xrcc4 single-mutant littermates (respectively [*106] 1.1 ± 0.2 SEM and 0.56 ± 0.15 SEM, p<0.0001, vs. 26.8 ± 2.6 SEM in thymus and 6.8 ± 1.2 SEM and 4.1 ± 1.0 SEM, p<0.0001, vs. 40.0 ± 3.8 SEM in spleen), and previously described Atm and Paxx single KO conditions (Abramowski et al., 2018). B cells expressing IgM were undetectable in BM from Xrcc4 mice and to a lesser extent in Xrcc4 mice (Figure 3C). B220+ B cells were strongly reduced compared to Xrcc4 or WT littermates in both settings (respectively 10.9% ± 1.4 SEM and 7.2% ± 0.96 SEM, p=0.001, vs. 21.7 ± 1.79 SEM in Xrcc4 and 26.3 ± 1.08 SEM in WT) (Figure 3D). Likewise, Xrcc4 and X4 mice presented with a substantial accumulation of DN3A thymocytes (93.4% ± 0.8 SEM and 98.2% ± 0.3 SEM, p=0.0001, respectively, vs. 76.1 ± 0.7 SEM in WT), which was significantly greater than the one observed for Xrcc4 (81.77 ± 1.4 SEM) or Nhej1-/- (81.1 ± 1.4 SEM) animals (Figure 3E and Abramowski et al., 2018). The immune phenotype of Xrcc4 and X4 mice is highly reminiscent of the SCID condition experienced by Nhej1-Paxx, Nhej1-Atm, and Nhej1-Rag2 DKO mice (Abramowski et al., 2018; Lescale et al., 2016a; Musilli et al., 2020; Zha et al., 2011) and suggested a possible impaired V(D)J recombination in these double mutants. Indeed, while Dβ2-Jβ2 (upper panel) and Vβ10-Dβ2Jβ2 (lower panel) rearrangements of the Tcrb locus were present in thymocytes from Xrcc4, Atm-/-, and WT mice (Figure 3F and G), as also previously observed in Paxx-/- mice (Abramowski et al., 2018), they were barely detectable in DNA from Xrcc4 and Xrcc4 thymus.

Figure 3.

PAXX and ATM are compensatory factors of X4M61R in immune development.

(A, B) Thymus and spleen cellularity of WT, Xrcc4, Xrcc4 and Xrcc4 6–9-week-old mice. (C) Immunostaining of BM and thymus cellular populations. Various relevant populations are highlighted in black gates. (D) Quantification of B220+ subpopulation of total BM. (E) Quantification of DN3A thymocyte subpopulation (DN3 CD28-CD25+) of total DN3 thymocytes (CD4-CD8-CD44-CD25+). (F) Schematic representation of PCR strategy to analyze Tcrb rearrangement according to Abramowski et al., 2018. (G) Autoradiogram of ladder of productive Dβ2-Jβ2 and Vβ10-Dβ2Jβ2 semi-quantitative PCR products revealed by the TCR-Jβ probe. Genomic DNA dilutions are represented with triangles. Germline allele configuration (GL) is revealed by the upper band of Dβ2-Jβ2 PCR. Southern were performed twice with two animals per genotype.

(A–C) Photographs of 7-week-old littermates.

We conclude that Xrcc4 crossed on Paxx-/- or Atm-/- phenocopy the SCID phenotype of Nhej1-Paxx and Nhej1-Atm DKO conditions owing to impaired V(D)J recombination. Therefore, aside from its essential role in Lig4 stabilization, X4 does participate in DNA coding-end tethering through its interaction with Xlf during V(D)J recombination, and PAXX/ATM are compensatory factors for this later function in vivo, as they are for Xlf. Alternatively, one may consider that the Xlf function being impaired in the context of X4M61R, PAXX/ATM complements this defect as already established.

Impaired brain development and immune deficiency in Xrcc4 mice

To further assess the function of X4M61R in vivo, we introduced Xrcc4 on Nhej1-/- background. No Xrcc4 double homozygous mice (hereafter named as Xrcc4) could be recovered out of 166 newborns obtained from various genotype combinations, when 23 were expected (Figure 4A), testifying the lethality of Xrcc4 embryos. Nevertheless, lived Xrcc4 fetuses could be recovered at E15.5 and E18.5. They were smaller than their Xrcc4+/- littermates, confirming abnormal development (Figure 4B and data not shown). This is reminiscent of the late embryonic lethality of Lig4-/-, Xrcc4-/-, and Nhej1-Paxx DKO mice (Abramowski et al., 2018; Balmus et al., 2016; Frank et al., 1998; Gao et al., 1998; Liu et al., 2017). Massive neuronal apoptosis was detected in E15.5 brain slices through cleaved-caspase 3 (CC3) staining in upper layers of dorsal telencephalon corresponding to the intermediate zone (IZ) and the cortical plate (CP) in Xrcc4 mice (Figure 4C–E). The IZ contains post-mitotic neurons migrating to populate the CP, while neural progenitors proliferate in lower layers: the ventricular zone (VZ) and sub-ventricular zone (SVZ). Thus, Xrcc4 embryonic lethality recapitulates, although to a lesser extent, the massive post-mitotic neuron apoptosis observed in Nhej1-Paxx DKO and Xrcc4-/- brains (Figure 4C and E), which causes the previously described embryonic lethality of these conditions. In contrast, Nhej1-/- and Xrcc4 single mutants showed modest increases of CC3 staining (Figure 4D and E), while we showed previously that Paxx-/- are indistinguishable from WT (Abramowski et al., 2018).

Figure 4.

Embryonic lethality and neuronal apoptosis of Xrcc4 embryos.

(A) Impaired development of Xrcc4 mice. Actual and expected numbers of Xrcc4 alive pups, E18.5 and E15.5 embryos. (B) Pictures of Xrcc4 and Xrcc4 E15.5 embryos. (C) Cleaved-caspase 3 (CC3, red) immunostaining of E15.5 brain slices. Scarce apoptotic cells are detected in the whole dorsal telencephalon from Nhej1-/- and Xrcc4 embryos. By contrast, massive neuronal apoptosis is observed in the upper layers of the developing cortex of Xrcc4, Xrcc4-/-, and Nhej1-Paxx DKO embryos. The white boxes indicate the location of the standard window (100 µm wide) spanning from the ventricular to pial surface used to quantify pyknotic nuclei in the dorsal telencephalons (see below). Scale bar: 100 µm. (D) Number of CC3-positive cells in the VZ/SVZ (black)—containing the cycling neural progenitors—and intermediate zone (IZ)/cortical plate (CP) (gray)—containing post-mitotic neurons—of the whole dorsal telencephalons from WT, Nhej1-/-, and Xrcc4 embryos. (E) Percentage of apoptotic (pyknotic) nuclei in the ventricular zone (VZ)/sub-ventricular zone (SVZ) (left panel) and the IZ/CP (right panel) found in the standard windows shown in (C). Data were obtained from both hemispheres of three embryos per condition.

We next analyzed the adaptive immune system development of Xrcc4 fetuses. B-lymphocyte development begins at E17.5 in the fetal liver (FL), with the rearrangement of the Igh locus and the intracellular expression of Ig-μH chains in CD19+ B220+ CD43+ Pro B cells, which is used as a proxy to evaluate Igh rearrangement completion (Kajikhina et al., 2016). Xrcc4 Pro-B cells failed to express intracellular μH chain when compared to WT (0.90% ± 0.09 SEM, p=0.008), which contrasted with the slight reduction observed in Xrcc4 and Nhej1-/- mice (respectively 16.6% ± 1.2 SEM and 19.1% ± 2.2 SEM vs. 32.8% ± 1.3 SEM, p=0.0016) (Figure 5A and B), arguing for a severe block in B cell maturation at Pro-B cell stage in Xrcc4 mice. Fetal thymocytes development of Xrcc4 and Nhej1-/- mice recapitulated young adult thymus phenotype with a modest block at DN3A stage compared to WT (respectively 83.3% ± 2.7 SEM and 86.2 ± 2.2 SEM vs. 70.3% ± 1.1 SEM, p=0.0004). In contrast, Xrcc4 thymocyte development was severely compromised from the DN3A stage on when compared to Xrcc4 and Nhej1-/- (99.4% ± 0.13 SEM, p=0.0007), and DP thymocytes or mature T cells were virtually absent from these thymii (Figure 5A and C). Dβ2-Jβ2 and Vβ10-Dβ2Jβ2 Tcrb genes rearrangements, which begin at around E13 (Ramond et al., 2014), were undetectable in Xrcc4 thymus, while readily observed in WT, Xrcc4, and Nhej1-/- mice (Figure 5D and E), further attesting for an acute V(D)J recombination defect in Xrcc4 cells.

Figure 5.

Complete V(D)J recombination defect in E18.5 Xrcc4 embryos.

(A) Immunostaining of E18.5 fetal CD19+ B220+ CD43+ hepatocytes and fetal thymocytes. Various relevant populations are highlighted in black gates. (B) Quantification of intracellular IgM expression in subpopulation of fetal Pro-B cells (CD19+ B220+ CD43+ surfaceIgM- hepatocytes). (C) Quantification of DN3A thymocyte subpopulation (DN3 CD28-CD25+) of total DN3 fetal thymocytes (CD4-CD8-CD44-CD25+). (D) Schematic representation of PCR strategy to analyze Tcrb rearrangement according to Abramowski et al., 2018. (E) Autoradiogram of ladder of productive Dβ2-Jβ2 and Vβ10-Dβ2Jβ2 direct PCR products revealed by the TCR-Jβ probe. Germline allele configuration (GL) is revealed by the upper band of Dβ2-Jβ2 PCR. Southern was performed once with two animal per genotype.

Altogether, Xrcc4 embryonic development phenocopy Nhej1-Paxx DKO, with a profound defect in V(D)J recombination and the apoptosis of post-mitotic neurons leading to embryonic lethality.

Discussion

We created an Xrcc4 separation of function allele in mice, which harbor the M61R substitution previously described to abolish X4-Xlf interaction (Ropars et al., 2011). The X4M61R protein, although weakly expressed, preserves the stabilization of Lig4 but severely compromises the NHEJ DNA repair function in affected mice. Nevertheless, Xrcc4 mice are viable, arguing that the embryonic lethality of Xrcc4 KO mice is most likely the result of the de facto absence of Lig4 in this setting, thus phenocopying Lig4 KO condition. Interestingly, this also indicates that the formation of the X4-Xlf filament, abrogated by the M61R mutation, is dispensable for post-mitotic neuron viability, the loss of which causes lethality in Xrcc4 and Lig4 KO settings.

Immune development of Xrcc4 mice phenocopied that of Nhej1-/- animals, with a slight decreased thymic cellularity, an increased thymocyte apoptosis, and skewed TCRα repertoire, which we previously linked to a suboptimal Tcra rearrangement efficiency (Berland et al., 2019; Roch et al., 2019; Vera et al., 2013). In addition to their TCRα repertoire bias, Xrcc4 thymocytes also faced a developmental delay at DN3A both in fetal and adult thymus like Nhej1 KO condition, further attesting for a suboptimal V(D)J recombination activity in these mice. Therefore, although X4 is required for V(D)J recombination through Lig4 stabilization, it is not mandatory for coding-ends tethering during V(D)J recombination. These results confirm that X4 exerts a Lig4-independent function through its interaction with Xlf that is decisive for the repair of genotoxic-induced DSBs but compensated for by other DNA repair factors, and possibly RAG2 itself as shown for Xlf (Lescale et al., 2016a), during V(D)J recombination. Xrcc4 and Xrcc4 double homozygous mice were severely immunocompromised, owing to aborted V(D)J recombination, as noticed in Nhej1-Paxx and Nhej1-Atm DKO. We conclude that PAXX and ATM are compensatory factors for X4M61R during V(D)J recombination. This result further supports our previously proposed model of ‘double DNA repair synapse’ in V(D)J recombination, mediated by the X4-Xlf and RAG2-PAXX-ATM axes, respectively (Abramowski et al., 2018; Betermier et al., 2020; Lescale et al., 2016a). In addition to their defect in V(D)J recombination, Xrcc4 double homozygous mice were embryonic lethal owing to massive post-mitotic neuron apoptosis, recapitulating Xrcc4-/- setting. This observation may be the result of the lower expression of the X4M61R protein, which could have additive/synthetic effects with the loss of Xlf interaction during NHEJ. Indeed, Nhej1 deficiency causes synthetic lethality with several other NHEJ-deficient conditions, such as Prkdc-/-, Mri-/-, H2ax-/-, and Paxx-/- (Abramowski et al., 2018; Hung et al., 2018; Xing et al., 2017; Zha et al., 2011). Therefore, Xlf can rescue very different NHEJ defects, through an unknown mechanism, perhaps in relation with recovery of replication fork stalling as shown in the Nhej1-H2ax DKO mice (Chen et al., 2019).

Altogether, the Xrcc4 separation of function allele highlights for the first time at least two independent roles of X4 in NHEJ and establishes that X4 is not mandatory for coding ends tethering during V(D)J recombination. This study also unravels novel interplays between X4, PAXX, ATM, and Xlf during development of the brain and the immune system.

Several deleterious mutations in the XRCC4 gene have been reported in humans (see de Villartay, 2015 for review), most of which are associated with microcephalic primordial dwarfism (MPD), gonadal failure, early-onset metabolic syndrome, and cardiomyopathies. The DNA repair deficiency in these patients is manifest when tested in vitro retrospectively. Most surprisingly however, they do not present noticeable signs of immune dysfunction. Altogether, the clinical presentation of these patients was not evocative of an impaired NHEJ given the known impact of its deficiency on the development of the adaptive immune system. Indeed, X4 mutations were not identified in these patients through hypothesis-driven candidate gene sequencing but rather through unsupervised whole-exome sequencing. Our present study now provides some hints as to explain the absence of immunological features in X4-deficient patients; when Lig4 expression is spared to some extent by hypomorphic mutations, allowing birth, XRCC4 appears not critically required for the development of the adaptive immune system.

Materials and methods

Mice

Nhej1-/- (Vera et al., 2013), Paxx-/- (Abramowski et al., 2018), Xrcc4-/- (Gao et al., 1998), and Atm-/- (Barlow et al., 1996) mice were maintained in pathogen-free environment. All experiments were performed in compliance with the French Ministry of Agriculture’s regulations for animal experiments (act 87847, 19 October 1987; modified in May 2001).

The Xrcc4 allele was generated through CRISPR/cas9 ribonucleoprotein (RNP) complex microinjection in C57BL/6J mouse zygotes pronuclei as described (Ucuncu et al., 2020). Xrcc4 exon 3-specific gRNA (5′-GTACCAGCTGCTCCTGACCC-3′) was designed using Crispor (http://crispor.tefor.net/oligonucleotide). The HR template was the ssODN ([Integrated DNA technologies, IDT] 5′-TCAAGGGAGAAATGCCGAGACTCCTTAGAAAAGAGGAACTTGTATGTACCAGCaGCACCGGAGCCTGGCACCAGTGCCTTTCTCAGCTCATCAATGTATTTTCCTTTCTCCCTAGCCATGTCATCAGCTTCTTGGGAAATCTCCAATTCAGAAACTATGGGAAAGATTAATTAAGGTGAT-3′; see Figure 1—figure supplement 1). F° chimera screen and Xrcc4 mice genotyping were obtained upon FspBI digestion of genomic DNA amplified with primers F 5′-AGGAGACGGAGGAAAAAGAGATG-3′ and R 5′-TACCCTCACAGAAACACAACTCA-3′.

Immunoblotting

Whole tissues or MEFs single-cell suspensions were washed in PBS. Cells were lysed in 50–200 μL of 50 mM Tris-HCl pH 7.5, 150 mM NaCl, 1% TritonX100, 1X phosphatase inhibitor cocktail 2 and 3 (ThermoFisher), and 1X protease inhibitor cocktail EDTA free (Roche) on ice for 45 min. Proteins migrated within Tris-Acrylamide gels and Tris-Glycine-SDS buffer and were transferred on PVDF membranes (Immobilon). Membranes were stained with the following antibodies: β-actin (mouse monoclonal AM4302, ThermoFisher), XRCC4 (goat monoclonal C-20, Santa Cruz), and DNA ligase IV (mouse monoclonal D-8, Santa Cruz). Membranes were revealed with appropriate secondary antibodies conjugated with infrared 700 or 800 dyes (LiCor). Immunoblotting was revealed by LiCor CLx analyzer and analyses were performed on Image Studio Lite software.

X4 transcript analysis

Total RNAs were obtained from thymocytes with Pure Like RNA mini kit (Invitrogen) and used for RT-PCR (SuperScript II reverse transcriptase; Invitrogen). X4 transcripts were amplified with primers F1 5′-TTGGGCGCATCGGTTTATCT-3′ R1 5′-GCTGCTAAGTTGAAAGCCTG-3′ F2 5′-TGCTTCTGAACCCAACGTACC-3′ R2 5′- AGGTGCTCGTTTTTGGCTTG-3′. F1-R1 PCR amplifies full-length transcript from 5′UTR to 3′UTR (1150 bp) and F2-R2 amplifies exon 2 to exon 4 (400 bp).

Flow cytometry analysis of cell populations

Cell phenotyping from 6- to 9-week-old mice was performed on thymus, spleen, and bone marrow using the following antibodies: CD4, CD8, CD25, CD28, CD44, CD69, B220, CD19, CD43, and IgM (all from Sony Biotechnology, using respectively PECy7, FITC, PerCPCy5.5, PE, BV510, APC, PE, PECy7, FITC, APC fluorophores). Intracellular IgM expression in E18.5 FL cells was performed as previously described (Abramowski et al., 2018) using CD19, B220, CD43, IgM for extracellular staining followed by cell fixation and permeabilization (Invitrogen) and intracellular IgM staining (all antibodies from Sony Biotechnology, using respectively PECy7, BV605, PE, FITC, and APC). Cells were recorded by FACS LSR-Fortessa X-20 and analyses were performed with FlowJo 10 software.

Cellular sensitivity to DSBs-inducing agents and thymocyte survival assay

For phleomycin sensitivity assay, 5000 MEF cells were seeded in triplicates and cultivated with increasing doses (0–300 ng/mL) of phleomycin. Living cells were counted by flow analysis with FACS LSR-Fortessa X-20 after 6 days of culture. Radiation sensitivity of T lymphocytes was performed as described (Abramowski et al., 2018). Ex vivo thymocyte survival assay was performed as described (Vera et al., 2013).

Quantitative real-time RT-PCR analysis

TaqMan PCR was performed on triplicates of 8 ng of reverse-transcribed RNA from freshly dissected total thymus as described (Vera et al., 2013).

Analysis of thymic TCR-alpha repertoire

Comprehensive TCR-alpha repertoire analyses were performed by 5′ Rapid amplification of complementary DNA (cDNA) ends (5′RACE) PCR/NGS (switching mechanism at the 5′end of the RNA transcript, SMARTα) from total thymus RNA as described (Abramowski et al., 2018; Roch et al., 2019; Vera et al., 2013). Sequencing data were analyzed with LymAnalyzer (Yu et al., 2016) to retrieve unique CDR3 clonotypes and determine T cell receptor alpha variable (TRAV) and T cell receptor alpha junction (TRAJ) gene segments. Frequencies of TRAV and TRAJ usage were implemented in principal component analysis (PCA) and hierarchical clustering on principal components (HCPC) analyses using the PCA() and HCPC() functions of the FactomineR package, respectively (http://factominer.free.fr/; Le et al., 2008) and graphics were generated using the Factoextra R package (http://www.sthda.com/english/rpkgs/factoextra; Kassambara, 2017).

TCRβ V(D)J recombination analysis

Tcrb rearrangements were analyzed by PCR on genomic DNA from total adult thymocytes or E18.5 fetal thymocytes as described (Abramowski et al., 2018).

E15.5 brain sections immunohistochemistry

Neuronal apoptosis was analyzed as previously described (Abramowski et al., 2018). E15.5 fetal heads were fixed overnight at 4°C by immersion in 4% paraformaldehyde and embedded in paraffin with a Tissu-tek processor (VIP, Leica). 5 μm coronal sections were then obtained using a microtome (Leica RM2125RT) and mounted onto glass slides for histological analyses. After paraffin removal and citrate treatment, the brain sections were permeabilized with 0.5% Triton X-100 in phosphate-buffered saline (PBS) for 15 min and incubated for 2 hr with 7.5% fetal bovine serum and 7.5% goat serum in PBS. The sections were incubated with rabbit anti-CC3 (Cell Signaling 9661) overnight at 4°C. After washing, the sections were incubated with goat anti-rabbit Alexa Fluor 488 or 594 conjugated secondary antibody (ThermoFisher) for 1 hr. After washing, nuclear staining was achieved by incubation with 4′-6-diamidino-2-phenylindole (DAPI) to quantify apoptosis induction by the detection of pyknotic nuclei (Roque et al., 2012). Slides were mounted under Fluoromount (Southern Biotechnologies Associates). Tissues were examined under a fluorescence microscope (50i, Nikon, Japan) with a 10× (NA = 0.3) objective in three channels (appearing red, green, and gray) as separate files. These images were then stacked with Photoshop software (Adobe).

Statistical analysis

Non-parametric Mann–Whitney test or two-tailed Fisher’s exact test was performed with α-risk = 0.05. p-Values were taken to be significant: *significant 0.05≥p>0.01; **very significant 0.01≥p>0.001; ***highly significant 0.001≥p>0.0001; ****highly significant p≤0.0001. Using G*Power software (https://www.psychologie.hhu.de/arbeitsgruppen/allgemeine-psychologie-und-arbeitspsychologie/gpower), we determined that a minimum of 47 newborns was necessary to conclude to an embryonic lethality of the Xrcc4 double KO with a power of 80% and a p-value < 0.05.

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Acceptance summary:

The novel mouse model described in this manuscript helps to not only clarify the role of XRCC4 in VDJ recombination and, thereby, for immune cell development but also reveals new functional interactions between XRCC4 and XLF that regulate neuronal apoptosis. The separation of function point mutation studied here demonstrates that the primary role of XRCC4 during VDJ recombination is to stabilize Lig4. This distinction also provides a ready explanation for earlier observations that humans with hypomorphic mutations in XRCC4 are not clinically immunodeficient. These insights will be broad interest to the immune development and DNA repair research communities.

Decision letter after peer review:

Thank you for submitting your article "A XRCC4 mutant mouse, a model for human X4 syndrome, reveals interplays with Xlf, PAXX, and ATM in lymphoid development" for consideration by eLife. Your article has been reviewed by 2 peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Satyajit Rath as the Senior Editor. The following individuals involved in review of your submission have agreed to reveal their identity: Michael Lieber (Reviewer #1); Karla Rodgers (Reviewer #2).

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

Essential revisions:

1) Provide a clearer and more detailed introduction to models of functions of XRCC4, XLF etc in end joining that highlights the fundamental question(s) being addressed via generation of the mouse model described in this manuscript.

2) Both reviewers list several inconsistencies in the figures, such as labeling of axes, percentages of cell populations, splenic cellularity in certain mutant mouse lines, scales in Figure 4, mean/standard deviations etc. these should be carefully checked and made uniform throughout the manuscript.

3) The protein levels of X4 and Lig4 in the double knock-out mice/embryos should be shown.

4) The Xlf-/- result is not shown in Figure 3E. Either the relevant citation needs to be included on this line, or the Xlf-/- result needs to be shown in Figure 3E.

The detailed comments from the reviewers are also appended below for providing context.

Reviewer #1 (Recommendations for the authors):

XRCC4 deficient mouse model exhibits late embryonic lethality and defective lymphogenesis. However, no clinically significant immunodeficiency was observed in human patients with XRCC4 (X4) hypomorphic mutations, even in those with no detectable X4 in the fibroblasts. The contradictory phenomena observed in mouse and human make the roles of X4 in lymphocyte development unclear. Roch et al., developed a unique X4 mutant mouse model and investigated the roles of X4 in V(D)J recombination. The authors first generated a viable X4M61R knock-in mouse model using CRISPR/Cas9 technique. The M61R mutation disrupts the interaction between X4 and XLF, but not the interaction of X4 and LIG4. The expression of X4 protein is severely inhibited in the homozygous X4 M61R/M61R (X4M61R) mice, while LIG4 can be partially stabilized by this mutant. The separation-of-function of this mutant could help identify the functions of Lig4 stabilizing activity of X4 and the roles of X4-XLF interaction. The authors further show that MEFs from X4M61R mice show mild sensitivity to IR and phleomycin, and X4M61R does not severely affect B- and T-lymphocyte development in mice. The authors therefore suggest that the X4-XLF interaction is not critical for V(D)J recombination in vivo. The authors further investigated the functional redundancies between PAXX, ATM, and X4 in lymphoid development by generating X4M61R-PAXX-/- and X4M61R-ATM-/- double mutant mice. The X4M61R-PAXX-/- and X4M61R-ATM-/- double mutant mice exhibit impaired V(D)J recombination, causing SCID phenotype. The authors further suggest that PAXX and ATM are compensatory factors of X4M61R in immune development. The authors also introduced X4M61R on XLF-/- background mice. They found that X4M61R-XLF-/- mice experience late embryonic lethality caused by massive neuronal apoptosis, and exhibit defect in V(D)J recombination.

1. In the Introduction section, the authors propose their idea of a "double DNA repair synapsis" model for repair of broken ends specifically caused by RAG1/2 complex during V(D)J recombination. However, the synapsis model is not clear here. The authors might want to provide more details about the model in this section, so readers could easily understand the model. The authors suggest that RAG2, PAXX, and ATM signaling mediate the first synapse to stabilize the RAG-cut broken ends; XRCC4-XLF filaments mediate the second synapse to tether the broken ends. The "two synapsis" model was proposed based on the observations of functional redundancy in V(D)J recombination between XLF and RAG2, PAXX, and ATM kinase activity. In the current manuscript, the authors seek to understand the contributions of XRCC4 to this model. However, the "two synapse" model is not very consistent with other studies. PAXX and XLF were reported to stabilize the synaptic complex of broken ends by interacting with Ku70/Ku80 complex (PMID: 29786079, PMID: 31399561, PMID: 33289484). This suggests that PAXX and XLF have overlapping functions in mediating end synapsis, and this does not mean PAXX and XLF are within different synaptic states (first synapse and second synapse). Moreover, if XRCC4 and XLF are in the same "second synapse" pathway, X4M61R-XLF-/- double mutant mice then should perform V(D)J recombination to some extent; this would lead to at least modest lymphocyte development, because the "first synapse" by either RAG2 or ATM, or PAXX could compensate for the "second synapse" as suggested by the authors. But X4M61R-XLF-/- mice reported by the authors are actually immune deficiency. The interaction of XLF and XRCC4 is indeed important for end synapsis, but it does not rely on the XRCC4-XLF filament. While the results here are interesting, the authors might need to rephrase the Introduction and frame the scientific question of this manuscript in a way that reconciles the above contradictions.

2. The authors conclude that PAXX/ATM are compensatory factors for X4 in immune development based on the SCID phenotype observed in X4M61R-PAXX-/- and X4M61R-ATM-/- double mutant mice. However, the results may also suggest that PAXX/ATM are compensatory factors for XLF but not directly for X4, because M61R mutation disrupts the interaction of X4 and XLF, which might result in XLF being unable to function in the X4M61R mice.

3. In lines 69-71, the authors cite Wang's paper to support the roles of XRCC4-XLF filaments in broken DNA end synapsis. However, in the Wang paper, XRCC4 and XLF are suggested to play roles mainly at the ends of the DNA (i.e., end-to-end synapsis) rather than through a XRCC4-XLF filament structure. Moreover, the statement of "X4 and PAXX binds DNA broken ends and stimulates X4-XLF DNA ends tethering in vitro" is not true. PAXX alone cannot bind dsDNA (as also reported in the cited Tadi et al., 2016 paper). It is true PAXX can stimulate end-to-end synapsis and ligation (PMID: 27705800, PMID: 31399561, PMID: 29786079). The stimulation by PAXX is through the interaction with Ku70/80, but not through the XRCC4-XLF filament.

4. The authors need to carefully examine their figures and related labels and legends to make sure they are correct. For example, one label is missing on Figure 1D; the x-axis labels of Figure 1G are incorrect; Figure 1 legend, line 134, "X4M61R vs X4-/- (bottom)" is inconsistent with those shown on Figure 1G.

5. The percentages of specific cell populations are shown on some flow cytometry panels but not on others. The ratios of the cell populations would make the flow cytometry data easier for readers to understand.

6. In line 164, the complete loss of viability was caused by 2 but not 1 Gy as shown on Figure 1G.

7. In line 228, the cellularity of X4M61R single mutant littermates in the spleen is missing.

8. The protein levels of X4 and Lig4 in the double knock-out mice/embryos could provide more information about the roles of X4 in lymphocyte development.

9. In the Introduction, the authors should also mention and consider the substantial literature supporting end-to-end synapsis models, which involve XLF interaction with XRCC4 and Ku. Both end-to-end synapsis and X4-XLF filament formation may be occurring.

10. The authors should consider that IR (e.g., 1 Gy) produces ~30 DSB per cell, whereas during V(D)J in T or B cells, there are likely to be only a few DSB per cell. One possible explanation for the difference between DNA repair roles and V(D)J is the number of events per cell and titration out of the XLF and X4 repair factors. Also, the chemistry and structure of the DNA ends for IR (and for phleomycin) are different than for V(D)J coding ends. The authors may want to consider that this could be an important reason for the difference in repair versus V(D)J for the role of X4.

Reviewer #2 (Recommendations for the authors):

XRCC4 (X4), a core component of the NHEJ DNA double strand break (DSB) repair pathway, functions to stabilize DNA Ligase IV (Lig4). X4 also associates with the XRCC4-like factor (Xlf) to form filaments that bridge and tether broken DNA ends. Both X4-/- and Lig4-/- mice show late embryonic lethality with massive neuronal apoptosis, as well as complete defects in V(D)J recombination. It was not known the extent that X4, aside from Lig4 stabilization, contributed to NHEJ in general DNA repair versus repair of programmed breaks during V(D)J recombination. To delineate differing functions of X4, a knock-in X4(M61R) mouse line was generated, where the M61R mutation was previously shown to disrupt interaction with Xlf while maintaining and stabilizing Lig4. A potential weakness of this model system is the reduced levels of X4(M61R) and Lig4 expression levels. Nevertheless, the overall results from this well-designed study are largely consistent with the authors' model and conclusions. Through combined genetic, immunological, and biochemical characterizations, the authors' illustrated three main points regarding X4 contributions to NHEJ. First, unlike X4-/-, X4(M61R) mice were viable with Lig4 stably expressed. Notably, while isolated X4(M61R) mouse cells showed deficiencies in repair of exogenously produced DNA DSBs, there were only relatively modest defects in B and T lymphocyte development, a marked contrast to that found with X4-/- mice. Second, The X4(M61R) mice phenocopy Xlf-/- mice by showing only minor defects in lymphocyte development, but with severe defects in V(D)J recombination when crossed with ATM or PAXX deficient mice. These results are consistent with the authors' model in which a "double DNA repair synapse" tethers the DNA ends in V(D)J recombination in a compensatory manner with one synapse mediated by the X4-Xlf filament and a separate synapse mediated by ATM signaling, PAXX, and the C-terminal region of RAG2. In a third point, X4(M61R) crossed with Xlf-/- mice showed late embryonic lethality due to post-mitotic neuronal apoptosis, with severe defects in V(D)J recombination in fetal hepatocytes and thymocytes. This latter point suggests that X4-Xlf filament cannot be compensated by a second DNA repair synapse in the fetal environment, perhaps due to the low expression of X4M61R. Alternatively, Xlf may function in an, as yet, unknown manner to rescue DNA repair defects of X4(M61R). In summary, using a novel separation of function X4 mutant, this study demonstrates that aside from its role in stabilizing Lig4, X4 is not requisite for B and T lymphocyte development likely due to functional redundancy of DNA end tethering in V(D)J recombination. Significantly, these results help to explain how patients with hypomorphic X4 mutations present with clear signs of DNA repair deficiency, but with no evident defects in development of the adaptive immune system.

This paper describes the development and characterization of a XRCC4 mutant mouse model, which helps to elucidate separate functions of X4 in NHEJ. Overall, this well-presented work demonstrates novel insights into the interplay between different NHEJ factors.

Essential revisions:

1) Provide a clearer and more detailed introduction to models of functions of XRCC4, XLF etc in end joining that highlights the fundamental question(s) being addressed via generation of the mouse model described in this manuscript.

The Introduction was rephrased and reorganized as requested. We added in particular 3 references of recently published articles describing the cryo-EM structure of the NHEJ core complex, thus improving our knowledge on DNA synapsis during NHEJ mediated DNA repair. We hope it increased the understanding of the scope of our study and readability of the manuscript.

2) Both reviewers list several inconsistencies in the figures, such as labeling of axes, percentages of cell populations, splenic cellularity in certain mutant mouse lines, scales in Figure 4, mean/standard deviations etc. these should be carefully checked and made uniform throughout the manuscript.

We carefully checked the Figures and Figure legends for mistakes and inconsistency.

3) The protein levels of X4 and Lig4 in the double knock-out mice/embryos should be shown.

Unfortunately, we will not be able to address this issue in a reasonable time frame. Because of the successive lockdown due to covid19 pandemic, we have been asked to sacrifice most of our living animals. Since the study was essentially finished and the manuscript was being prepared we had no choice but kill the mice related to this project.

4) The Xlf-/- result is not shown in Figure 3E. Either the relevant citation needs to be included on this line, or the Xlf-/- result needs to be shown in Figure 3E.

Citation has been added.

The detailed comments from the reviewers are also appended below for providing context.

Reviewer #1:

XRCC4 deficient mouse model exhibits late embryonic lethality and defective lymphogenesis. However, no clinically significant immunodeficiency was observed in human patients with XRCC4 (X4) hypomorphic mutations, even in those with no detectable X4 in the fibroblasts. The contradictory phenomena observed in mouse and human make the roles of X4 in lymphocyte development unclear. Roch et al., developed a unique X4 mutant mouse model and investigated the roles of X4 in V(D)J recombination. The authors first generated a viable X4M61R knock-in mouse model using CRISPR/Cas9 technique. The M61R mutation disrupts the interaction between X4 and XLF, but not the interaction of X4 and LIG4. The expression of X4 protein is severely inhibited in the homozygous X4 M61R/M61R (X4M61R) mice, while LIG4 can be partially stabilized by this mutant. The separation-of-function of this mutant could help identify the functions of Lig4 stabilizing activity of X4 and the roles of X4-XLF interaction. The authors further show that MEFs from X4M61R mice show mild sensitivity to IR and phleomycin, and X4M61R does not severely affect B- and T-lymphocyte development in mice. The authors therefore suggest that the X4-XLF interaction is not critical for V(D)J recombination in vivo. The authors further investigated the functional redundancies between PAXX, ATM, and X4 in lymphoid development by generating X4M61R-PAXX-/- and X4M61R-ATM-/- double mutant mice. The X4M61R-PAXX-/- and X4M61R-ATM-/- double mutant mice exhibit impaired V(D)J recombination, causing SCID phenotype. The authors further suggest that PAXX and ATM are compensatory factors of X4M61R in immune development. The authors also introduced X4M61R on XLF-/- background mice. They found that X4M61R-XLF-/- mice experience late embryonic lethality caused by massive neuronal apoptosis, and exhibit defect in V(D)J recombination.

1. In the Introduction section, the authors propose their idea of a "double DNA repair synapsis" model for repair of broken ends specifically caused by RAG1/2 complex during V(D)J recombination. However, the synapsis model is not clear here. The authors might want to provide more details about the model in this section, so readers could easily understand the model.

The Introduction was rephrased and reorganized to take into account these comment. We hope it increased the understanding of the scope of our study and readability of the manuscript.

The authors suggest that RAG2, PAXX, and ATM signaling mediate the first synapse to stabilize the RAG-cut broken ends; XRCC4-XLF filaments mediate the second synapse to tether the broken ends. The "two synapsis" model was proposed based on the observations of functional redundancy in V(D)J recombination between XLF and RAG2, PAXX, and ATM kinase activity. In the current manuscript, the authors seek to understand the contributions of XRCC4 to this model. However, the "two synapse" model is not very consistent with other studies.

As we mentioned in the introduction, the "two synapse" model is a strict specificity of the V(D)J recombination process (as opposed to random genotoxic-induced DSBs) because of the presence of the R1/2 post cleavage complex remaining on broken DNA ends as we proposed in our previous papers (Lescale et al., Nat. Com. 2016 and Betermier et al., TCB 2020). We believe it's the reason why it may appear sometimes inconsistent with purely DNA repair studies performed in vitro or in cellulo following genotoxics. The most striking inconsistency being the fact that although Xlf is a critical NHEJ factor as shown by the impressive radiosensitivity of Xlf deficient cells, the consequence on V(D)J recombination is almost null in contrast to what was well established for most NHEJ deficiencies.

PAXX and XLF were reported to stabilize the synaptic complex of broken ends by interacting with Ku70/Ku80 complex (PMID: 29786079, PMID: 31399561, PMID: 33289484). This suggests that PAXX and XLF have overlapping functions in mediating end synapsis, and this does not mean PAXX and XLF are within different synaptic states (first synapse and second synapse). Moreover, if XRCC4 and XLF are in the same "second synapse" pathway, X4M61R-XLF-/- double mutant mice then should perform V(D)J recombination to some extent; this would lead to at least modest lymphocyte development, because the "first synapse" by either RAG2 or ATM, or PAXX could compensate for the "second synapse" as suggested by the authors. But X4M61R-XLF-/- mice reported by the authors are actually immune deficiency. The interaction of XLF and XRCC4 is indeed important for end synapsis, but it does not rely on the XRCC4-XLF filament. While the results here are interesting, the authors might need to rephrase the Introduction and frame the scientific question of this manuscript in a way that reconciles the above contradictions.

The Introduction was rephrased and reorganized. We hope it increased the understanding of the scope of our study and readability of the manuscript.

2. The authors conclude that PAXX/ATM are compensatory factors for X4 in immune development based on the SCID phenotype observed in X4M61R-PAXX-/- and X4M61R-ATM-/- double mutant mice. However, the results may also suggest that PAXX/ATM are compensatory factors for XLF but not directly for X4, because M61R mutation disrupts the interaction of X4 and XLF, which might result in XLF being unable to function in the X4M61R mice.

We thank the Reviewer for this thought and added a sentence accordingly.

3. In lines 69-71, the authors cite Wang's paper to support the roles of XRCC4-XLF filaments in broken DNA end synapsis. However, in the Wang paper, XRCC4 and XLF are suggested to play roles mainly at the ends of the DNA (i.e., end-to-end synapsis) rather than through a XRCC4-XLF filament structure. Moreover, the statement of "X4 and PAXX binds DNA broken ends and stimulates X4-XLF DNA ends tethering in vitro" is not true. PAXX alone cannot bind dsDNA (as also reported in the cited Tadi et al., 2016 paper). It is true PAXX can stimulate end-to-end synapsis and ligation (PMID: 27705800, PMID: 31399561, PMID: 29786079). The stimulation by PAXX is through the interaction with Ku70/80, but not through the XRCC4-XLF filament.

The Introduction was rephrased and reorganized to take into account these comment. We hope it increased the understanding of the scope of our study and readability of the manuscript.

4. The authors need to carefully examine their figures and related labels and legends to make sure they are correct. For example, one label is missing on Figure 1D; the x-axis labels of Figure 1G are incorrect; Figure 1 legend, line 134, "X4M61R vs X4-/- (bottom)" is inconsistent with those shown on Figure 1G.

We carefully checked the Figures and Figure legends for mistakes and inconsistency.

5. The percentages of specific cell populations are shown on some flow cytometry panels but not on others. The ratios of the cell populations would make the flow cytometry data easier for readers to understand.

The % of cell population was shown only on Fig2D. For consistency with all other FACS panel we removed the % values on Fig2D. The actual values of gated cell population are computed on side histograms.

As suggested, we tried representation of the data by ratio rather than true values. As shown in Author response image 1 on two examples, we believe that the ratio representation (right) makes the data less readable in addition to losing the information of the real numbers (left) as it compresses the differences. We decided to keep data representation as it was.

Author response image 1.

6. In line 164, the complete loss of viability was caused by 2 but not 1 Gy as shown on Figure 1G.

Corrected.

7. In line 228, the cellularity of X4M61R single mutant littermates in the spleen is missing.

Mean+/-SEM were checked for consistency.

8. The protein levels of X4 and Lig4 in the double knock-out mice/embryos could provide more information about the roles of X4 in lymphocyte development.

Unfortunately, we will not be able to address this issue in a reasonable time frame. Because of the successive lockdown due to covid19 pandemic, we have been asked to sacrifice most of our living animals. Since the study was essentially finished and the manuscript was being prepared we had no choice but kill the mice related to this project.

9. In the Introduction, the authors should also mention and consider the substantial literature supporting end-to-end synapsis models, which involve XLF interaction with XRCC4 and Ku. Both end-to-end synapsis and X4-XLF filament formation may be occurring.

We added a paragraph on DNA end synapsis with reference to Zhao et al., for further detailed reading and the 3 references on resolution of NHEJ core complexes through cryo-EM.

10. The authors should consider that IR (e.g., 1 Gy) produces ~30 DSB per cell, whereas during V(D)J in T or B cells, there are likely to be only a few DSB per cell. One possible explanation for the difference between DNA repair roles and V(D)J is the number of events per cell and titration out of the XLF and X4 repair factors. Also, the chemistry and structure of the DNA ends for IR (and for phleomycin) are different than for V(D)J coding ends. The authors may want to consider that this could be an important reason for the difference in repair versus V(D)J for the role of X4.

We agree that there are major qualitative and quantitative differences beetween VDJ and IR generated DSBs. Nevertheless, again taking Xlf deficiency as an example the one V(D)J break is repaired in Xlf KO mice but not anymore in many Xlf/NHEJ (or RAG2cc) double KO conditions. So we do not think that the number of breaks or their nature is the central issue, but rather the existence of a V(D)J specific backup system for the repair of prDSB.